Appendix 1: Worksheets for

Practice on SPSS by
Workshop Participants

Appendix 1:Worksheets For Practice on SPSS By Workshop Participants – Update July 26, 2001
WORKSHEET 1

Summarising Numerical Data in SPSS

Open the Excel spreadsheet file with the Paterson Inlet blue cod data called PICodSize.
(DME CHCRO-28728)

1. Copy data (not labels) from columns B to H into SPSS, that is, data in columns
with the headings:

Site Summary
Number Year Month Reserve n Mean SD

2. Open SPSS and paste data into the data editor window. Go into variable view
(see bottom of screen), under name insert appropriate variable names, e.g., label
the first variable site number.

3. Summarise the site-summary variables: n, mean and sd using Analyse >

Which plot is the most useful for displaying the data?

4. In the output window double click on the histogram for the variable "n" to open
the chart editor. Under the option Chart click on axis and change the axis of the
graph so that it has a maximum at 20. Add a normal curve to the graph (Chart,
Options and Display Normal Curve).

5. Repeat the steps above but where the sites within the proposed reserve are
separated from sites outside the proposed reserve. The Factor List is the
variable that was called "Reserve" in the spreadsheet (1 = in the proposed
reserve, 0 = outside the proposed reserve). This should produce box plots on the
single graph for each variable.

Is there evidence of differences in fish in sites inside and outside the proposed
reserve?

How would you test this statistically?

6. The data were collected over 4 years (1994 to 1998). Summarise the mean fish
length at survey sites by year and by survey site status (i.e., whether it is inside

Appendix 1:Worksheets For Practice on SPSS By Workshop Participants – Update July 26, 2001
 or outside the proposed reserve). One way to do this is to use Analyse >
Compare Means > Means. Put the variable "Mean" in the Dependent List.
The Independent List Variables is "Year". Click on Next and make "Reserve"
the variable for the next layer.

Plot the data summarised for year and reserve status as box plots, e.g., Graphs >
Boxplot > Clustered > Data in Charts - are for Groups of Cases.

Do the differences between mean fish length at sites within and outside the
proposed reserve differ over time?

Is there anything unusual about the data in 1997? Recall there was little
difference between the average of the fish means in 1997 for sites within and
outside the proposed reserve however, the box plots show the median of the data.

Appendix 1:Worksheets For Practice on SPSS By Workshop Participants – Update July 26, 2001
WORKSHEET 2

SUMMARISING NUMERICAL DATA IN SPSS CONTINUED

Open the Excel spreadsheet file Dataset7 (DME CHCRO-28729), tree conditions at
Ryan Creek, Heaphy Valley in 1995 and 1999. The purpose of this exercise is to
compare the measures on trees for 1995 and 1999, and to look at the relationships
between some of these measures.

The measures that will be considered are Folcov, Dbtop, Dbwho, Brtop and Brwho.
These are given tree by tree in the file and it is these values that will be considered in
this worksheet. Note however that for statistical analysis purposes the results for
several trees on the same plot may not be independent.

See the definitions of the variables in the spreadsheet (or in the printed copy of page 1
of the data).

1. Copy and paste the values of Year, Line, Plot, Folcov, Dbtop, Dbwho, Brtop,
Brwho and Stuse into SPSS Data editor window. Go into variable view (see
bottom of screen), under name insert appropriate variable names, e.g., label the
first variable year.

2. Summarise the results for the six variables Folcov to Brwho using Analyse >
Descriptive Statistics > Frequencies. Look carefully at the various options,
and make sure that you print a histogram for each of these variables.

3. A useful device is the Matrix Plot, which produces on one page plots of several
variables each against each other. Use Graphs > Scatter > Matrix to produce
one of these plots for the six variables. One of the options here is to use a
different colour for different categories of cases. Use this option to produce
matrix plots with different symbols for 1995 and 1999, and then for different
lines.

4. The matrix plot with all six variables is a bit too cramped. Try producing a
series of matrix graphs to answer the following questions:

(a) How are Dbtop and Dbwho related, and is this the same for both years
and all lines?

(b) How are Brtop and Brwho related, and is this the same for both years and
all lines?
Appendix 1:Worksheets For Practice on SPSS By Workshop Participants – Update July 26, 2001
 (c) How are Dbwho, Brwho and Stuse related, are does this vary with the
year or line?

5. Check that you can copy and paste graphs and text from the SPSS output
window into MS Word.

Appendix 1:Worksheets For Practice on SPSS By Workshop Participants – Update July 26, 2001
WORKSHEET 3

FACTORIAL ANALYSIS OF VARIANCE

In Module 4, Section 4.5, there was a three factor analysis of variance was carried out
using the General Linear Model > univariate option in SPSS. The dependent
variable was the average number of birds of all species recorded in 40 5-minute counts.
The factors Area (treated or control), Time (before or after poisoning), and summer
(1995/96 to 1989/99) were all assumed to be fixed in this analysis. The data are in the
Excel file DATSET12.xls. (DME CHCRO-28730).

For this worksheet we see how the analysis changes if the two areas are regarded as
being randomly selected from a population of possible areas that could be used, and
how multiple comparison tests can be used to compare factor levels.

Open the Excel spreadsheet file Dataset12 (DME CHCRO-28730)

1. Copy and paste the data into SPSS Data editor window. Go into variable view
(see bottom of screen), under name insert appropriate variable names, e.g., label
the first variable area.

2. Choose the options Analyse > General Linear Model > univariate. Choose
mcount as the dependent variable, and let Summer, Time and Area have fixed
effects (i.e., assume that the levels of these factors in the data are all the levels of
interest). Click on the Model button. Choose the full factorial if this is not set
already.

3. Run the analysis. You should get the analysis of variance table that is part of
Table 4.6 in Module 4.

4. Return to the data window and then back to the general factorial option. Change
the Area factor to one with random effects, and run the analysis. Note how the
analysis of variance table is now much more complicated. In particular, the error
term used for F-tests now changes according to what is being tested. This
demonstrates the importance of making the correct assumptions about fixed and
random effects once you have more than one factor for your data.

5. Return to the data window and then the general factorial option again. Either
leave Area as a random effects factor, or make it fixed again. You can now
check the effects of other options in the analysis. There are many of these and

Appendix 1:Worksheets For Practice on SPSS By Workshop Participants – Update July 26, 2001
 the help facility will guide you through the ones that you choose to look at. The
following notes may help.

(a) With Model you can choose to include only some of the interactions in
your model, using Custom. This may be appropriate if the full factorial
analysis indicates that some interactions are insignificant because the
estimates of the important interactions should be improved if the
insignificant ones are removed. Note, however, that you should not
remove an interaction while leaving a higher order interaction that
includes it still in the model. For example, if the factors are A, B and C,
do not remove A.B while leaving A.B.C in. Why is this?

(b) With Contrasts you can choose to compare the means at different levels
of a factor. See the SPSS help for more information about the alternative
ways of doing this. For real data you might, for example have the
situation where the first level of a factor is a control and the other levels
are increasing doses. You may then wish to compare levels 2, 3, 4 etc.
with the first level.

(c) With Plots you can plot the mean levels for one factor against the levels
for a second factor. Such plots indicate the type of interaction that exists,
if any.

(d) With Post Hoc there are 18 different ways to compare the means for
different factor levels, allowing for multiple testing. Browse the SPSS
help facility for recommendations about which of these to use.

(e) Covariates can be added into the model. For example, if individual
animals are the sample units, and large animals are expected to have
higher levels for the response variable, irrespective of factor effects, then
putting a measure of size in as a covariate makes an adjustment for the
sizes of the animals used in the study. This is done by adding the term
;LQWRWKHPRGHOZKHUH;LVVL]H2IFRXUVH\RXPXVWKDYHDFROXPQ
of sizes in your data sheet to be able to do this.

(f) The WLS option allows the observations in the analysis to have different
weights. A common reason for this is that the observations are actually
mean values for samples of different sizes. In that case it is appropriate to
put the sample size in as the weight variable.

Appendix 1:Worksheets For Practice on SPSS By Workshop Participants – Update July 26, 2001
 (g) Finally, the Options button gives you the ability to print out mean values
for different factor levels and combinations of factor levels, test for
unequal variances at different factor levels, get residual plots, etc. Again,
consult the SPSS help facility for more information.

Appendix 1:Worksheets For Practice on SPSS By Workshop Participants – Update July 26, 2001