Two-way ANOVA

Page 1 of 28

Two-way ANOVA

You measured a response to three different drugs in both

men and women. Is the response affected by drug? By
gender? Are the two intertwined? These are the kinds of
questions that two-way ANOVA answers.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Key concepts: Two-way ANOVA

Two-way ANOVA, also called two-factor ANOVA, determines how a

response is affected by two factors. For example, you might measure a

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 2 of 28

response to three different drugs in both men and women. In this

example, drug treatment is one factor and gender is the other.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Entering two-way ANOVA data

Groups are defined by rows and columns
Prism organizes data for two-way ANOVA differently than do most other programs.
Prism does not use grouping variables. Instead, use rows and columns to designate the different
groups (levels) of each factor. Each data set (column) represents a different level of one factor,
and each row represents a different level of the other factor.
From the Welcome or New table dialog, choose the Grouped tab to set up the right kind of table.

Two ways to enter the data

Before entering data, choose which grouping variable is defined by the rows, and which is defined
by the columns. For example, if you are comparing men and women at three time points, there are
two ways to organize the data:

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 3 of 28

Of course, the ANOVA results will be identical no matter which way you enter the data. But the
choice does matter, as it influences how the graph will appear and how Prism can do multiple
comparison post tests.

Your choice affects how the graph will look

When you create a graph of the data, the points or bars from each column can have a different
appearance and color.
If you enter data as shown in the first approach above, men and women will appear in bars of
different color, with three bars of each color representing the three time points. This is illustrated
in the left panel below.
The right panel below shows a graph of the second data table shown above. There is one bar color
and fill for Before, another for During, and another for After. Men and Women appear as two bars
of identical appearance.

Prism's multiple comparison post tests only compare within a row

Prism can only perform post tests within a row, comparing columns. Using the first approach,
Prism compares Men vs. Women at each time point. Using the second approach, Prism compares
Before vs. During, Before vs. After, and During vs. After for each gender. Arrange your data so
these post tests address your scientific questions.

Use the transpose analysis to change your mind

What happens if after entering and analyzing your data using one of the choices above, you then
realize you wish you had done it the other way? You don't need to reenter your data. Instead use

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 4 of 28

Prism's transpose analysis, and then do two-way ANOVA on the results.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Choosing two-way ANOVA

What is two-way ANOVA used for?
Two-way ANOVA, also called two-factor ANOVA, determines how a response is affected by two
factors. For example, you might measure a response to three different drugs in both men and
women. In this example, drug treatment is one factor and gender is the other.
Two-way ANOVA simultaneously asks three questions:
1. Does the first factor systematically affect the results? In our example: Are the mean responses
the same for all three drugs?
2. Does the second factor systematically affect the results? In our example: Are the mean
responses the same for men and women?
3. Do the two factors interact? In our example: Are the differences between drugs the same for
men and women? Or equivalently, is the difference between men and women the same for all
drugs?
Although the outcome measure (dependent variable) is a continuous variable, each factor must be
categorical, for example: male or female; low, medium or high dose; or wild type or mutant.
ANOVA is not an appropriate test for assessing the effects of a continuous variable, such as blood
pressure or hormone level (use a regression technique instead).

Two-way ANOVA choices

Repeated measures
Choose a repeated-measures analysis when the experiment used paired or matched subjects.
Prism can calculate repeated-measures two-way ANOVA with matching by either row or column,
but not both. Details. This is sometimes called a mixed model.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 5 of 28

You can only choose repeated measures when you have entered two or more side-by-side values
into subcolumns for each row and dataset.

Post tests
Following two-way ANOVA, there are many possible multiple comparison tests that can help you
focus in on what is really going on. However, Prism performs only the post tests biologists use
most frequently. At each row, Prism will compare each column with every column, or compare
every column with a control column.

Variable names
If you enter names for the factors that define the rows and columns, the results will be easier to
follow.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Point of confusion: ANOVA with a quantitative factor

Two-way ANOVA is sometimes used when one of the factors is quantitative, such as when
comparing time courses or dose response curves. In these situations one of the factors is dose or
time.
GraphPad's advice: If one of your factors is
quantitative (such as time or dose) think hard before
choosing two-way ANOVA. Other analyses may make
more sense.

Interpreting P values with a quantitative factor

Let's imagine you compare two treatments at six time points. The ANOVA analysis treats different
time points exactly as it would treat different drugs or different species. The concept of trend is
entirely ignored (except in some special post tests).
The two-way ANOVA will report three P values:
One P value tests the null hypothesis that time has no effect on the outcome. It rarely makes
sense to test this hypothesis. Of course time affects the outcome! That's why you did a time
course.
Another P value tests the null hypothesis that the treatment makes no difference, on average.
This can be somewhat useful. But you probably expect no difference at early (or maybe late)
time points, and only care about differences at late time points. So it may not be useful to ask
if, on average, the treatments differ.
The third P value tests for interaction. The null hypothesis is that any difference between
treatments is identical at all time points. But if you collect data at time zero, or at early time
points, you don't expect to find any difference then. Your experiment really is designed to ask
about later time points. In this situation, you expect an interaction, so finding a small P value
for interaction does not help you understand your data.
ANOVA pays no attention to the order of your time points (or doses). If you randomly scramble the
time points or doses, two-way ANOVA would report identical results. In other words, ANOVA
ignores the entire point of the experiment, when one of the factors is quantitative.

Interpreting post tests with a quantitative factor

What about post tests?
Some scientists like to ask which is the lowest dose (or time) at which the change in response is
statistically significant. Post tests can give you the answer, but the answer depends on sample
size. Run more subjects, or more doses or time points for each curve, and the answer will change.
With a large enough sample size (at each dose), you will find that a tiny dose causes a statistically
significant, but biologically trivial, effect. This kind of analysis does not ask a fundamental
question, and so the results are rarely helpful.
If you want to know the minimally effective dose, consider finding the minimum dose that causes

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 6 of 28

an effect bigger than some threshold you set based on physiology. For example, find the minimum
dose that raises the pulse rate by more than 10 beats per minute.
If you look at all the post tests (and not just ask which is the lowest dose or time point that gives a
'significant' effect), you can get results that make no sense. You might find that the difference is
significant at time points 3, 5, 6 and 9 but not at time points 1, 2, 4, 7, 8 and 10. How do you
interpret that? Knowing at which doses or time points the treatment had a statistically significant
rarely helps you understand the biology of the system and rarely helps you design new
experiments.

Alternatives to two-way ANOVA

What is the alternative to two-way ANOVA?
If you have a repeated measures design, consider using an alternative to ANOVA. Will G Hopkins
calls the alternative within-subject modeling.
First, quantify the data for each subject in some biologically meaningful way. Perhaps this would
be the area under the curve. Perhaps the peak level. Perhaps the time to peak. Perhaps you can fit
a curve and determine a rate constant or a slope.
Now take these values (the areas or rate constants...) and compare between groups of subjects
using a t test (if two treatments) or one-way ANOVA (if three or more). Unlike two-way ANOVA,
this kind of analysis follows the scientific logic of the experiment, and so leads to results that are
understandable and can lead you to the next step (designing a better experiment).
If you don't have a repeated measures design, you can still fit a curve for each treatment. Then
compare slopes, or EC50s, or lag times as part of the linear or nonlinear regression.
Think hard about what your scientific goals are, and try to find a way to make the statistical
testing match the scientific goals. In many cases, you'll find a better approach than using two-way
ANOVA.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Q&A: Two-way ANOVA

I know the mean, SD (or SEM) and sample size for each group. Which tests can I run?
You can enter data as mean, SD (or SEM) and N, and Prism can compute two-way ANOVA. It is not
possible to compute repeated measures ANOVA without access to the raw data.

I only know the group means, and don't have the raw data and don't know their SD or SEM. Can I
run ANOVA?
No. ANOVA compares the difference among group means with the scatter within the groups, taking
into account sample size. If you only know the means, there is no possible way to do any statistical
comparison.

I want to compare three groups. The outcome has two possibilities, and I know the fraction of
each possible outcome in each group. How can I compare the groups?
Not with ANOVA. Enter your data into a contingency table and analyze with a chi-square test.

What does 'two-way' mean?

Two-way ANOVA, also called two-factor ANOVA, determines how a response is affected by two
factors. For example, you might measure a response to three different drugs at two time points.
The two factors are drug and time.
If you measure response to three different drugs at two time points with subjects from two age
ranges, then you have three factors: drug, time and age. Prism does not perform three-way
ANOVA, but other programs do.

What does 'repeated measures' mean? How is it different than 'randomized block'?
The term repeated-measures strictly applies only when you give treatments repeatedly to each
subject, and the term randomized block is used when you randomly assign treatments within each

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 7 of 28

group (block) of matched subjects. The analyses are identical for repeated-measures and
randomized block experiments, and Prism always uses the term repeated-measures.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Ordinary (not repeated measures) two-way ANOVA

Two-way ANOVA, also called two-factor ANOVA, determines how a

response is affected by two factors. For example, you might measure a
response to three different drugs in both men and women. In this
example, drug treatment is one factor and gender is the other. If each
subject was only assessed once (different subjects got different drugs),
and the males and females are not paired in some way (spouses,
brother-sister), then you have ordinary (not repeated measures) twoway ANOVA.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 8 of 28

How to: Two-way ANOVA

Two-way ANOVA, also called two-factor ANOVA, determines how a response is affected by two
factors. For example, you might measure a response to three different drugs in both men and
women. Drug treatment is one factor and gender is the other.
Prism uses a unique way to enter data. You use rows and columns to designate the different
groups (levels) of each factor. Each data set (column) represents a different level of one factor,
and each row represents a different level of the other factor. You need to decide which factor is
defined by rows, and which by columns. Your choice won't affect the ANOVA results, but the choice
is important as it affects the appearance of graphs and the kinds of post tests Prism can compare.
This page explains how to do ordinary two-way ANOVA. If you want to do repeated measures twoway ANOVA see separate examples for repeated measures by row and repeated measures by
column.

1. Create a data table

From the Welcome (or New Data Table and Graph) dialog, choose the Grouped tab.

Entering raw data

If you are not ready to enter your own data, chose to use sample data and choose: Two-way
ANOVA -- Ordinary.
If you plan to enter your own data, it is important that you choose the subcolumn format correctly,
for the maximum number of replicates you have.
Leave the graph set to its default -- interleaved bars, vertical.

Entering averaged data

If you have already averaged your replicates in another program, you can choose (at the bottom of
the dialog) to enter and plot the mean and SD (or SEM) and N. If your data has more than 52
replicates, this is the only way to enter data into Prism for two-way ANOVA.

2. Enter data
Here are the sample data:

Note that one value is blank. It is fine to have some missing values, but you must have at least
one value in each row for each data set. The following table cannot be analyzed by two-way
ANOVA because there are no data for treated women. But it doesn't matter much that there are
only two (not three) replicates for control men and treated men.

If you are entering mean, SD (or SEM) and N, You must enter N (number of replicates) but it is ok
if N is not always the same.

3. Choose two-way ANOVA

1. From the data table, click

on the toolbar.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 9 of 28

2. Choose Two-way ANOVA from the list of grouped analyses.

3. Select 'no matching' because your data doesn't involve repeated measure. Choose post tests if
they will help you interpret your results, and enter the name of the factors that define columns
and rows.

5. Interpret the results

Interpreting results: Two-way ANOVA
Graphing tips: Two-way ANOVA data
How Prism computes two-way ANOVA
Checklist: Two-way ANOVA

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Interpreting results: Two-way ANOVA

Two-way ANOVA determines how a response is affected by two factors. For example, you might
measure a response to three different drugs in both men and women.

ANOVA table
The ANOVA table breaks down the overall variability between measurements (expressed as the
sum of squares) into four components:
Interactions between row and column. These are differences between rows that are not the
same at each column, equivalent to variation between columns that is not the same at each
row.
Variability among columns.
Variability among rows.
Residual or error. Variation among replicates not related to systematic differences between
rows and columns.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 10 of 28

The ANOVA table shows how the sum of squares is partitioned into the four components. Most
scientists will skip these results, which are not especially informative unless you have studied
statistics in depth. For each component, the table shows sum-of-squares, degrees of freedom,
mean square, and the F ratio. Each F ratio is the ratio of the mean-square value for that source of
variation to the residual mean square (with repeated-measures ANOVA, the denominator of one F
ratio is the mean square for matching rather than residual mean square). If the null hypothesis is
true, the F ratio is likely to be close to 1.0. If the null hypothesis is not true, the F ratio is likely to
be greater than 1.0. The F ratios are not very informative by themselves, but are used to
determine P values.

P values
Two-way ANOVA partitions the overall variance of the outcome variable into three components,
plus a residual (or error) term. Therefore it computes P values that test three null hypotheses
(repeated measures two-way ANOVA adds yet another P value).

Interaction P value
The null hypothesis is that there is no interaction between columns (data sets) and rows. More
precisely, the null hypothesis states that any systematic differences between columns are the
same for each row and that any systematic differences between rows are the same for each
column. Often the test of interaction is the most important of the three tests. If columns represent
drugs and rows represent gender, then the null hypothesis is that the differences between the
drugs are consistent for men and women.
The P value answers this question:
If the null hypothesis is true, what is the chance of randomly sampling subjects and ending up
with as much (or more) interaction than you have observed?
The graph on the left below shows no interaction. The treatment has about the same effect in
males and females. The graph on the right, in contrast, shows a huge interaction. the effect of the
treatment is completely different in males (treatment increases the concentration) and females
(where the treatment decreases the concentration). In this example, the treatment effect goes in
the opposite direction for males and females. But the test for interaction does not test whether the
effect goes in different directions. It tests whether the average treatment effect is the same for
each row (each gender, for this example).

Testing for interaction requires that you enter replicate values or mean and SD (or SEM) and N. If
you entered only a single value for each row/column pair, Prism assumes that there is no
interaction, and continues with the other calculations. Depending on your experimental design, this
assumption may or may not make sense.
Tip: If the interaction is statistically significant, you
won't learn much from the other two P values. Instead
focus on the multiple comparison post tests.

Column factor P value

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 11 of 28

The null hypothesis is that the mean of each column (totally ignoring the rows) is the same in the
overall population, and that all differences we see between column means are due to chance. In
the example graphed above, results for control and treated were entered in different columns (with
males and females being entered in different rows). The null hypothesis is that the treatment was
ineffective so control and treated values differ only due to chance. The P value answers this
question: If the null hypothesis is true, what is the chance of randomly obtaining column means as
different (or more so) than you have observed?
In the example shown in the left graph above, the P value for the column factor (treatment) is
0.0002. The treatment has an effect that is statistically significant.
In the example shown in the right graph above, the P value for the column factor (treatment) is
very high (0.54). On average, the treatment effect is indistinguishable from random variation. But
this P value is not meaningful in this example. Since the interaction P value is low, you know that
the effect of the treatment is not the same at each row (each gender, for this example). In fact,
for this example, the treatment has opposite effects in males and females. Accordingly, asking
about the overall, average, treatment effect doesn't make any sense.

Row factor P value

The null hypothesis is that the mean of each row (totally ignoring the columns) is the same in the
overall population, and that all differences we see between row means are due to chance. In the
example above, the rows represent gender, so the null hypothesis is that the mean response is the
same for men and women. The P value answers this question: If the null hypothesis is true, what
is the chance of randomly obtaining row means as different (or more so) than you have observed?
In both examples above, the P value for the row factor (gender) is very low.

Post tests
Prism can compare columns at each row using multiple comparison post tests.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Graphing tips: Two-way ANOVA

The graph above shows three ways to plot the sample data for two-way ANOVA.
The graphs on the left and middle interleave the data sets. This is set on the second tab of the
Format Graphs dialog. In this case, the data sets are defined by the figure legend, and the groups
(rows) are defined by the labels on the X axis.
The graph on the right has the data sets grouped. In this graph, the labels on the X axis show the
row title -- one per bar. You can use the "number format" choice in the Format Axes dialog to
change this to Column titles -- one per set of bars. With this choice, there wouldn't be much point

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 12 of 28

in also having the legend shown in the box, and you would need to define the side by side bars
("serum starved" vs "normal culture" for this example) in the figure legend.
The graph on the left has the appearance set as a column dot plot. The other two graphs have the
appearance set as bars with error bars plotted from the mean and SD. I prefer the column dot plot
as it shows all the data, without taking up more space and without being harder to interpret.
Don't forget to include in the figure legend whether the error bars are SD or SEM or something
different.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

How Prism computes two-way ANOVA

Two-way ANOVA calculations are quite standard, and these comments only discuss some of the
ambiguities.

Model I (fixed effects) vs. Model II (random effects) ANOVA

To understand the difference between fixed and random factors, consider an example of comparing
responses in three species at three times. If you were interested in those three particular species,
then species is considered to be a fixed factor. It would be a random factor if you were interested
in differences between species in general, and you randomly selected those three species. Time is
considered to be a fixed factor if you chose time points to span the interval you are interested in.
Time would be a random factor if you picked those three time points at random. Since this is not
likely, time is almost always considered to be a fixed factor.
When both row and column variables are fixed factors, the analysis is called Model I ANOVA. When
both row and column variables are random factors, the analysis is called Model II ANOVA. When
one is random and one is fixed, it is termed mixed effects (Model III) ANOVA. Prism calculates only
Model I two-way ANOVA. Since most experiments deal with fixed-factor variables, this is rarely a
limitation.

Missing values
If some values are missing, two-way ANOVA calculations are challenging. Prism uses the method
detailed in SA Glantz and BK Slinker, Primer of Applied Regression and Analysis of Variance,
McGraw-Hill, 1990. This method converts the ANOVA problem to a multiple regression problem and
then displays the results as ANOVA. Prism performs multiple regression three times each time
presenting columns, rows, and interaction to the multiple regression procedure in a different order.
Although it calculates each sum-of-squares three times, Prism only displays the sum-of-squares for
the factor entered last into the multiple regression equation. These are called Type III sum-ofsquares.
Prism cannot perform repeated-measures two-way ANOVA if any values are missing. It is OK to
have different numbers of numbers of subjects in each group, so long as you have complete data
(at each time point or dose) for each subject.

Data entered as mean, N and SD (or SEM)

If your data are balanced (same sample size for each condition), you'll get the same results if you
enter raw data, or if you enter mean, SD (or SEM), and N. If your data are unbalanced, it is
impossible to calculate precise results from data entered as mean, SD (or SEM), and N. Instead,
Prism uses a simpler method called analysis of unweighted means. This method is detailed in LD
Fisher and G vanBelle, Biostatistics, John Wiley, 1993. If sample size is the same in all groups, and
in some other special cases, this simpler method gives exactly the same results as obtained by
analysis of the raw data. In other cases, however, the results will only be approximately correct. If
your data are almost balanced (just one or a few missing values), the approximation is a good
one. When data are unbalanced, you should enter individual replicates whenever possible.

Single values without replicates

Prism can perform two-way ANOVA even if you have entered only a single replicate for each
column/row pair. This kind of data does not let you test for interaction between rows and columns
(random variability and interaction can't be distinguished unless you measure replicates). Instead,
Prism assumes that there is no interaction and only tests for row and column effects. If this

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 13 of 28

assumption is not valid, then the P values for row and column effects won't be meaningful.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Analysis checklist: Two-way ANOVA

Two-way ANOVA, also called two-factor ANOVA,
determines how a response is affected by two factors. For
example, you might measure a response to three different
drugs in both men and women. In this example, drug
treatment is one factor and gender is the other. Read
elsewhere to learn about choosing a test, and interpreting
the results. Then review this analysis checklist before
accepting the results.
.

Are the populations distributed according to a Gaussian distribution?

Two-way ANOVA assumes that your replicates are sampled from Gaussian distributions. While
this assumption is not too important with large samples, it is important with small sample sizes,
especially with unequal sample sizes. Prism does not test for violations of this assumption. If you
really don't think your data are sampled from a Gaussian distribution (and no transform will
make the distribution Gaussian), you should consider performing nonparametric two-way
ANOVA. Prism does not offer this test.
ANOVA also assumes that all sets of replicates have the same SD overall, and that any
differences between SDs are due to random sampling.

Are the data unmatched?

Standard two-way ANOVA works by comparing the differences among group means with the
pooled standard deviations of the groups. If the data are matched, then you should choose
repeated-measures ANOVA instead. If the matching is effective in controlling for experimental
variability, repeated-measures ANOVA will be more powerful than regular ANOVA.

Are the errors independent?

The term error refers to the difference between each value and the mean of all the replicates.
The results of two-way ANOVA only make sense when the scatter is random that whatever
factor caused a value to be too high or too low affects only that one value. Prism cannot test this
assumption. You must think about the experimental design. For example, the errors are not
independent if you have six replicates, but these were obtained from two animals in triplicate. In
this case, some factor may cause all values from one animal to be high or low.

Do you really want to compare means?

Two-way ANOVA compares the means. It is possible to have a tiny P value clear evidence that
the population means are different even if the distributions overlap considerably. In some
situations for example, assessing the usefulness of a diagnostic test you may be more
interested in the overlap of the distributions than in differences between means.

Are there two factors?

One-way ANOVA compares three or more groups defined by one factor. For example, you might
compare a control group with a drug treatment group and a group treated with drug plus
antagonist. Or you might compare a control group with five different drug treatments. Prism has
a separate analysis for one-way ANOVA.
Some experiments involve more than two factors. For example, you might compare three
different drugs in men and women at four time points. There are three factors in that
experiment: drug treatment, gender and time. These data need to be analyzed by three-way

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 14 of 28

ANOVA, also called three-factor ANOVA. Prism does not perform three-way ANOVA.

Are both factors fixed rather than random?

Prism performs Type I ANOVA, also known as fixed-effect ANOVA. This tests for differences
among the means of the particular groups you have collected data from. Different calculations
are needed if you randomly selected groups from an infinite (or at least large) number of
possible groups, and want to reach conclusions about differences among ALL the groups, even
the ones you didn't include in this experiment.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Repeated measures two-way ANOVA

Two-way ANOVA, also called two-factor ANOVA, determines how a

response is affected by two factors. For example, you might measure a
response to three different drugs in both men and women. In this
example, drug treatment is one factor and gender is the other. If each
subject was assessed repeatedly (each time after treatment with a
different drug), then you have repeated measures two-way ANOVA

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 15 of 28

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

What is repeated measures?

You should choose a repeated-measures analysis when the experiment used paired or matched
subjects. Prism can calculate repeated-measures two-way ANOVA with matching by either row or
column, but not both. This is sometimes called a mixed model.

The table above shows example data testing the effects of three doses of drugs in control and
treated animals. The decision to use repeated measures depends on the experimental design.

Repeated measures by row

Here is an experimental design that would require analysis using repeated measures by row:
The experiment was done with six animals, two for each dose. The control values were
measured first in all six animals. Then you applied a treatment to all the animals and made the
measurement again. In the table above, the value at row 1, column A, Y1 (23) came from the
same animal as the value at row 1, column B, Y1 (28). The matching is by row.

Repeated measures by column

Here is an experimental design that would require analysis using repeated measures by column:
The experiment was done with four animals. First each animal was exposed to a treatment (or
placebo). After measuring the baseline data (dose=zero), you inject the first dose and make the
measurement again. Then inject the second dose and measure again. The values in the first Y1
column (23, 34, and 43) were repeated measurements from the same animal. The other three
subcolumns came from three other animals. The matching was by column.

Repeated measures in both factors

Here is an experimental design where both factors are repeated measures:
The experiment was done with two animals. First you measured the baseline (control, zero
dose). Then you injected dose 1 and made the next measurement, then dose 2 and measured
again. Then you gave the animal the experimental treatment, waited an appropriate period of
time, and made the three measurements again. Finally, you repeated the experiment with
another animal (Y2). So a single animal provided data from both Y1 subcolumns (23, 34, 43
and 28, 41, 56).
Prism cannot perform two-way ANOVA with repeated measures in both directions, and so cannot
analyze the data from this experiment.

"Repeated measures" vs. "randomized block" experiments

The term repeated measures is appropriate when you made repeated measurements from each
subject.
Some experiments involve matching but not repeated measurements. The term randomizedblock describes these kinds of experiments. For example, imagine that the three rows were three
different cell lines. All the Y1 data came from one experiment, and all the Y2 data came from
another experiment performed a month later. The value at row 1, column A, Y1 (23) and the value
at row 1, column B, Y1 (28) came from the same experiment (same cell passage, same reagents).
The matching is by row.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 16 of 28

Randomized block data are analyzed identically to repeated-measures data. Prism always uses the
term repeated measures, so you should choose repeated measures analyses when your
experiment follows a randomized block design.

Example without repeated measures

Finally, here is an example of an experiment done with replicates but no repeated measures:
The experiment was done with six animals. Each animal was given one of two treatments at one
of three doses. The measurement was then made in duplicate. The value at row 1, column A, Y1
(23) came from the same animal as the value at row 1, column A, Y2 (24). Since the matching is
within a treatment group, it is a replicate, not a repeated measure. Analyze these data with
ordinary two-way ANOVA, not repeated-measures ANOVA.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

How to: Two-way ANOVA. Repeated measures by row

Two-way ANOVA, also called two-factor ANOVA, determines how a response is affected by two
factors. For example, you might measure a response to three different drugs in both men and
women. Drug treatment is one factor and gender is the other.
Prism uses a unique way to enter data. You use rows and columns to designate the different
groups (levels) of each factor. Each data set (column) represents a different level of one factor,
and each row represents a different level of the other factor. You need to decide which factor is
defined by rows, and which by columns. Your choice will not affect the ANOVA results, but the
choice is important as it affects the appearance of graphs and the kinds of post tests Prism can
compare.
This page shows you how to enter and analyze data with repeated measurements placed on a
single row. Use a different 'how to' page if you enter repeated measurements in a subcolumn.

1. Create a data table and enter data

From the Welcome (or New Data Table and Graph) dialog, choose the Grouped tab.
If you are not ready to enter your own data, chose to use sample data and choose: Two-way
ANOVA data -- RM by rows. ("RM" means Repeated Measures).
If you plan to enter your own data, it is important that you choose the subcolumn format correctly,
for the maximum number of replicates you have.
Since your data are repeated measures, you want to make a graph that shows that. Choose the
fifth choice on the second row of graph types, so values are connected properly on the graph. If
you choose sample data, Prism automatically chooses the appropriate graph.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 17 of 28

Enter the data putting matched values in the same row, in corresponding columns. In the sample
data shown below, the two circled values represent repeated measurements in one animal.

2. Choose two-way ANOVA

1. From the data table, click

on the toolbar.

2. Choose Two-way ANOVA from the list of grouped analyses.

3. Choose the option that specifies that matched values are spread across a row.

3. Interpret the results

Interpreting results: Repeated measures two-way ANOVA
Graphing tips: Repeated measures two-way ANOVA

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 18 of 28

Checklist: Repeated measures two-way ANOVA

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

How to: Two-way ANOVA. Repeated measures by column

Two-way ANOVA, also called two-factor ANOVA, determines how a response is affected by two
factors. For example, you might measure a response to three different drugs in both men and
women. Drug treatment is one factor and gender is the other.
Prism uses a unique way to enter data. You use rows and columns to designate the different
groups (levels) of each factor. Each data set (column) represents a different level of one factor,
and each row represents a different level of the other factor. You need to decide which factor is
defined by rows, and which by columns. Your choice will not affect the ANOVA results, but the
choice is important as it affects the appearance of graphs and the kinds of post tests Prism can
compare.
This page shows you how to enter and analyze data with repeated measurements placed in a
subcolumn. Use a different 'how to' page if you enter repeated measurements in a row.

1. Create a data table and enter data

From the Welcome (or New Data Table and Graph) dialog, choose the Grouped tab.
If you are not ready to enter your own data, chose to use sample data and choose: Two-way
ANOVA data -- RM by columns. ("RM" means Repeated Measures).
If you plan to enter your own data, it is important that you choose the subcolumn format correctly,
for the maximum number of subjects you have with any treatment.
Since your data are repeated measures, you want to make a graph that shows that. Choose the
second choice on the second row of graph types, so values are connected properly on the graph.
Choose to plot each replicate, connecting each subcolumn. If you choose the sample data, Prism
will automatically choose the appropriate graph.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 19 of 28

Arrange your data so the data sets (columns) represent different levels of one factor, and different
rows represent different levels of the other factor. The sample data set compares five time points
in two subjects under control conditions and two after treatment.

Each subcolumn (one is marked with a red arrow above) represents repeated measurements on a
single subject.

Missing values
It is OK if some treatments have more subjects than others. In this case, some subcolumns will be
entirely blank. But you must enter a measurement for each row for each subject. Prism cannot
handle missing measurements with repeated measures ANOVA.

2. Choose two-way ANOVA

1. From the data table, click

on the toolbar.

2. Choose Two-way ANOVA from the list of grouped analyses.

3. Choose the option that specifies that matched values are stacked in a subcolumn.

Also, choose post tests if they will help you interpret your results, and enter the name of the
factors that define columns and rows.

3. Interpret the results

Interpreting results: Repeated measures two-way ANOVA
Graphing tips: Repeated measures two-way ANOVA
Checklist: Repeated measures two-way ANOVA

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Graphing tips: Repeated measures two-way ANOVA

Graphing two-way ANOVA with repeated measures by row
From the welcome dialog, you can choose a graph designed for repeated measures by rows. This is
the fifth choice on the bottom row of graphs in the two-way tab.
Customize the graph within the Format Graph dialog:

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 20 of 28

The appearance (for all data sets) should be 'Before-After'.

Plot either symbols and lines or lines only. Choose the latter if you want to plot arrows.
The line style drop down lets you choose arrow heads.

Graphing two-way ANOVA with repeated measures by column

From the welcome dialog, you can choose a graph designed for repeated measures by rows. This is
the second choice on the bottom row of graphs in the two-way tab.
Customize the graph within the Format Graph dialog:

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 21 of 28

The appearance (for all data sets) should be "Each replicate".

if you plot the replicates as 'Staggered", Prism will move them right or left to prevent overlap.
In this example, none of the points overlap so 'Staggered' and 'Aligned' look the same.
Check the option to plot 'one line for each subcolumn'.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 22 of 28

Interpreting results: Repeated measures two-way ANOVA

Are you sure that ANOVA is the best analysis?
Before interpreting the ANOVA results, first do a reality check. If one of the factors is a
quantitative factor like time or dose, consider alternatives to ANOVA.

Interpreting P values from repeated measures two-way ANOVA

When interpreting the results of two-way ANOVA, most of the considerations are the same whether
or not you have repeated measures. So read the general page on interpreting two-way ANOVA
results first.
Repeated measures ANOVA reports an additional P value: the P value for subject (matching) This
tests the null hypothesis that the matching was not effective. You expect a low P value if the
repeated-measures design was effective in controlling for variability between subjects. If the P
value was high, reconsider your decision to use repeated-measures ANOVA

How the repeated measures ANOVA is calculated

Prism computes repeated-measures two-way ANOVA calculations using the standard method
explained especially well in SA Glantz and BK Slinker, Primer of Applied Regression and Analysis of
Variance, McGraw-Hill, 1990.

Interpreting post tests after repeated measures ANOVA

The use of post tests after repeated measures ANOVA is somewhat controversial.
The post tests performed by Prism, use the mean square residual for all comparisons. This is a
pooled value that assess variability in all the groups. If you assume that variability really is the
same in all groups (with any differences due to chance) this gives you more power. This makes
sense, as you get to use data from all time points to assess variability, even when comparing only
two times. Prism does this automatically.
But with repeated measures, it is common that the scatter increases with time, so later treatments
give a more variable response than earlier treatments. In this case, some argue that the post tests
are misleading.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Analysis checklist: Repeated measures two-way ANOVA

Two-way ANOVA, also called two-factor ANOVA,
determines how a response is affected by two factors.
"Repeated measures" means that one of the factors was
repeated. For example you might compare two treatments,
and measure each subject at four time points (repeated).
Read elsewhere to learn about choosing a test, graphing
the data, and interpreting the results. Then review this
analysis checklist before accepting the results.

Are the data matched?

If the matching is effective in controlling for experimental variability, repeated-measures ANOVA
will be more powerful than regular ANOVA.

Are there two factors?

One-way ANOVA compares three or more groups defined by one factor. For example, you might
compare a control group with a drug treatment group and a group treated with drug plus
antagonist. Or you might compare a control group with five different drug treatments. Prism has
a separate analysis for one-way ANOVA.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 23 of 28

Some experiments involve more than two factors. For example, you might compare three
different drugs in men and women at four time points. There are three factors in that
experiment: drug treatment, gender and time. These data need to be analyzed by three-way
ANOVA, also called three-factor ANOVA. Prism does not perform three-way ANOVA.

Are both factors fixed rather than random?

Prism performs Type I ANOVA, also known as fixed-effect ANOVA. This tests for differences
among the means of the particular groups you have collected data from. Different calculations
are needed if you randomly selected groups from an infinite (or at least large) number of
possible groups, and want to reach conclusions about differences among ALL the groups, even
the ones you didn't include in this experiment.

Can you accept the assumption of 'circularity'

A random factor that causes a measurement in one subject to be a bit high (or low) should have
no affect on the next measurement in the same subject. This assumption is called circularity or
sphericity. It is closely related to another term you may encounter in advanced texts,
compound symmetry.
You only have to worry about the assumption of circularity when your experiment truly is a
repeated-measures experiment, with measurements from a single subject. You don't have to
worry about circularity with randomized block experiments where you used a matched set of
subjects (or a matched set of experiments)
Repeated-measures ANOVA is quite sensitive to violations of the assumption of circularity. If the
assumption is violated, the P value will be too low. You'll violate this assumption when the
repeated measurements are made too close together so that random factors that cause a
particular value to be high (or low) don't wash away or dissipate before the next measurement.
To avoid violating the assumption, wait long enough between treatments so the subject is
essentially the same as before the treatment. Also randomize the order of treatments, when
possible.
Some advanced programs do special calculations to detect, and correct for, violations of the
circularity or sphericity assumption. Prism does not.

Consider alternatives to repeated measures two-way ANOVA.

Two-way ANOVA may not answer the questions your experiment was designed to address.
Consider alternatives.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Post tests after two-way ANOVA

After two-way ANOVA, you may wish to perform post tests to narrow
down where differences are significant. Prism can perform some of
these analyses.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 24 of 28

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

What post tests can Prism perform following two-way

ANOVA?
Following two-way ANOVA, there are many possible multiple comparison tests that can help you
focus in on what is really going on. However, Prism performs only the post tests described below
(which biologists use most frequently).
Prism can only compare one value with another value within the same row.
Prism does not compare values within the same column.
Prism does not compute row means and then compare one row mean with another.
Prism does not compute column means and then compare one column mean with another.
Post tests are most often done when there are two columns of data (representing control and
treated), and multiple rows (perhaps representing different time points). Prism can perform post
tests to compare the control value and the treated value at each time.
If you have three columns, Prism can also perform post tests. Say that data set A is control, data
set B is one treatment, and data set C is another treatment. Each row represents a different time
point. Prism can do two kinds of post tests. It can do all possible comparisons at each time point
(row). In this example, there are three columns, and prism can compare A with B, A with C, and B
with C at each row. Or you can specify that one data set is the control (A in this case) and Prism
will compare each other data set to the control. In this example, Prism would compare A with B,
and A with C at each time point (row).

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

How Prism calculates post tests after two-way ANOVA

Prism performs post tests following two-way ANOVA using the Bonferroni method as detailed in
pages 741-744 and 771 in J Neter, W Wasserman, and MH Kutner, Applied Linear Statistical
Models, 3rd edition, Irwin, 1990.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 25 of 28

The numerator is the difference between the mean response in the two data sets (usually control
and treated) at a particular row (usually dose or time point). The denominator combines the
number of replicates in the two groups at that dose with the mean square of the residuals
(sometimes called the mean square of the error), which is a pooled measure of variability at all
doses.
Statistical significance is determined by comparing the t ratio with the t distribution for the number
of df shown in the ANOVA table for MSresidual, applying the Bonferroni correction for multiple
comparisons. The Bonferroni correction lowers the P value that you consider to be significant to
0.05 divided by the number of comparisons. This means that if you have five rows of data with two
columns, the P value has to be less than 0.05/5, or 0.01, for any particular row in order to be
considered significant with P<0.05. This correction ensures that the 5% probability applies to the
entire family of comparisons, and not separately to each individual comparison.
Post tests following repeated-measures two-way ANOVA use exactly the same equation if the
repeated measures are by row. If the repeated measures are by column, the term (SSsubject +
SSresidual)/(DFsubject + DFresidual) is used instead of MSresidual in the equation above, and the
number of degrees of freedom equals the sum of DFsubject and DFresidual.

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

Calculating more general post tests

Need for more general post tests
Prism only performs one kind of post test following two-way ANOVA. If your experimental situation
requires different post tests, you can calculate them by hand without too much trouble. Or use the
free web QuickCalcs provided at graphpad.com.
Consider this example where you measure a response to a drug after treatment with vehicle,
agonist, or agonist+antagonist, in both men and women.

Prism will compare the two columns at each row. For this example, Prism's built-in post tests
compare the two columns at each row, thus asking:
Do the control responses differ between men and women?
Do the agonist-stimulated responses differ between men and women?
Do the response in the presence of both agonist and antagonist differ between men and
women?
If these questions match your experimental aims, Prism's built-in post tests will suffice. Many
biological experiments compare two responses at several time points or doses, and Prism built-in
post tests are just what you need for these experiments. But you may wish to perform different
post tests. In this example, based on the experimental design above, you might want to ask these
questions:
For men, is the agonist-stimulated response different than control? (Did the agonist work?)
For women, is the agonist-stimulated response different than control?
For men, is the agonist response different than the response in the presence of agonist plus
antagonist? (Did the antagonist work?)
For women, is the agonist response different than the response in the presence of agonist plus
antagonist?
For men, is the response in the presence of agonist plus antagonist different than control?

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 26 of 28

(Does the antagonist completely block agonist response?)

For women, is the response in the presence of agonist plus antagonist different than control?
One could imagine making many more comparisons, but we'll make just these six. The fewer
comparisons you make, the more power you'll have to find differences, so it is important to focus
on the comparisons that make the most sense. But you must choose the comparisons based on
experimental design and the questions you care about. Ideally you should pick the comparisons
before you see the data. It is not appropriate to choose the comparisons you are interested in after
seeing the data.

How to do the calculations

For each comparison (post test) you want to know:
What is the 95% confidence interval for the difference?
Is the difference statistically significant (P<0.05)?
Although Prism won't calculate these values for you, you can easily do the calculations yourself,
starting from Prism's ANOVA table. For each comparison, calculate the confidence interval for the
difference between means using this equation (from pages 741-744 and 771, J Neter, W
Wasserman, and MH Kutner, Applied Linear Statistical Models, 3rd edition, Irwin, 1990).

In this equation, mean1 and mean 1 are the means of the two groups you are comparing, and N1
and N2 are their sample size. MSresidual is reported in the ANOVA results, and is the same for all
post tests.
The variable t* is the critical value from the student t distribution, using the Bonferroni correction
for multiple corrections. When making a single confidence interval, t* is the value of the t ratio
that corresponds to a two-tail P value of 0.05 (or whatever significance level you chose). If you are
making six comparisons, t* is the t ratio that corresponds to a P value of 0.05/6, or 0.00833. Find
the value using this Excel formula =TINV(0.00833,6), which equals 3.863. The first parameter is
the significance level corrected for multiple comparisons; the second is the number of degrees of
freedom for the ANOVA (residuals for regular two-way ANOVA, subject' for repeated measures).
The value of t* will be the same for each comparison. Its value depends on the degree of
confidence you desire, the number of degrees of freedom in the ANOVA, and the number of
comparisons you made.
To determine significance levels, calculate for each comparison:

The variables are the same as those used in the confidence interval calculations, but notice the key
difference. Here, you calculate a t ratio for each comparison, and then use it to determine the
significance level (as explained in the next paragraph). When computing a confidence interval, you
choose a confidence level (95% is standard) and use that to determine a fixed value from the t
distribution, which we call t*. Note that the numerator is the absolute value of the difference
between means, so the t ratio will always be positive.
To determine the significance level, compare the values of the t ratio computed for each
comparison against the standard values, which we abbreviate t*. For example, to determine
whether the comparison is significant at the 5% level (P<0.05), compare the t ratios computed for
each comparison to the t* value calculated for a confidence interval of 95% (equivalent to a
significance level of 5%, or a P value of 0.05) corrected for the number of comparisons and taking

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 27 of 28

into account the number of degrees of freedom. As shown above, this value is 3.863. If a t ratio is
greater than t*, then that comparison is significant at the 5% significance level. To determine
whether a comparison is significant at the stricter 1% level, calculate the t ratio corresponding to a
confidence interval of 99% (P value of 0.01) with six comparisons and six degrees of freedom. First
divide 0.01 by 6 (number of comparisons), which is 0.001667. Then use the Excel formula =TINV
(0.001667,6) to find the critical t ratio of 5.398. Each comparison that has a t ratio greater than
5.398 is significant at the 1% level.
Tip: All these calculations can be performed using a free
QuickCalcs web calculator.

Example
For this example, here are the values you need to do the calculations (or enter into the web
calculator).
Comparison
1: Men. Agonist vs.
control
2: Women. Agonist vs.
control
3: Men. Agonist vs.
Ag+Ant
4: Women. Agonist vs.
Ag+Ant
5: Men Control vs.
Ag+Ant
6: Women. Control vs.
Ag+Ant

Mean1
176.0

Mean2
98.5

N1
2

N2
2

206.5

100.0

176.0

116.0

206.5

121.0

98.5

116.0

100.0

121.0

And here are the results:

Comparison
1:
2:
3:
4:
5:
6:

Men. Agonist vs. control

Women. Agonist vs. control
Men. Agonist vs. Ag+Ant
Women. Agonist vs. Ag+Ant
Men Control vs. Ag+Ant
Women. Control vs. Ag+Ant

Comparison
1: Men. Agonist vs. control
2: Women. Agonist vs. control
3: Men. Agonist vs. Ag+Ant
4: Women. Agonist vs. Ag+Ant
5: Men Control vs. Ag+Ant
6: Women Control vs. Ag+Ant

Significant? (P <
0.05?)
Yes
Yes
Yes
Yes
No
No
Mean1 - Mean2
+ 77.5
+ 106.5
+ 60.0
+ 85.5
-17.5
-21.0

t
8.747
12.020
6.772
9.650
1.975
2.370
95% CI of difference
+ 43.3 to + 111.7
+ 72.3 to + 140.7
+ 25.8 to + 94.2
+ 51.3 to + 119.7
-51.7 to + 16.7
-55.2 to + 13.2

The calculations account for multiple comparisons. This means that the 95% confidence level
applies to all the confidence intervals. You can be 95% sure that all the intervals include the true
value. The 95% probability applies to the entire family of confidence intervals, not to each
individual interval. Similarly, if the null hypothesis were true (that all groups really have the same
mean, and all observed differences are due to chance) there will be a 95% chance that all
comparisons will be not significant, and a 5% chance that any one or more of the comparisons will
be deemed statistically significant with P< 0.05.
For the sample data, we conclude that the agonist increases the response in both men and women.
The combination of antagonist plus agonist decreases the response down to a level that is
indistinguishable from the control response.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009

Two-way ANOVA

Page 28 of 28

Copyright (c) 2007 GraphPad Software Inc. All rights reserved.

file://C:\Documents and Settings\jmpf\Configuracin local\Temp\~hhEFC4.htm

11/18/2009