Scribd
Upload
48 views4 pages

Functions of Covariates Reduce Error Variance

ANCOVA can serve two main purposes: 1. Reduce error variance by explaining previously unexplained variance in the dependent variable, leading to a more sensitive measure of the experimental effect. 2. Eliminate confounds by accounting for variables known to influence the dependent variable and removing their biasing effect. ANCOVA makes adjustments to dependent variable values to remove the influence of measured covariates, and then performs an ANOVA on the adjusted values. It has the standard parametric assumptions plus two additional requirements: independence of covariates and treatment, and homogeneity of regression slopes.

Uploaded by

Deepa Menon
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
48 views4 pages

Functions of Covariates Reduce Error Variance

ANCOVA can serve two main purposes: 1. Reduce error variance by explaining previously unexplained variance in the dependent variable, leading to a more sensitive measure of the experimental effect. 2. Eliminate confounds by accounting for variables known to influence the dependent variable and removing their biasing effect. ANCOVA makes adjustments to dependent variable values to remove the influence of measured covariates, and then performs an ANOVA on the adjusted values. It has the standard parametric assumptions plus two additional requirements: independence of covariates and treatment, and homogeneity of regression slopes.

Uploaded by

Deepa Menon
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 4

Functions of Covariates

Reduce error variance

– The F-ratio in ANOVA compares the amount of variability explained by the experimental
manipulation (MSM), against the variability that it cannot explain (MS R). What we hope
is that the covariate explains some of the variance that was previously unexplained.
This has the effect of reducing the unexplained variance (i.e., MS R becomes smaller) and
so our F-ratio gets bigger. In real terms this means we obtain a more sensitive measure
of the experimental effect.

Eliminate Confounds

– In any experiment, there may be unmeasured variables that confound the results (i.e., a
variable that varies systematically with the experimental manipulation). If any variables
are known to influence the dependent variable being measured, then ANCOVA can help
to remove the bias of these variables. Once a possible confounding variable has been
identified, it can be measured and entered into the analysis as a covariate.

ANCOVA is like an ANOVA on the values of the DV, after removing the influence of the covariate, rather
than on the original values.

ANCOVA has the same basic assumptions of all of the parametric tests, but also has two additional
requirements:

1. independence of the covariate and treatment effect.

2. homogeneity of regression slopes.

We assume the covariate is correlated with the outcome variable (DV), so that as scores on the
covariate change, scores on the DV change by a similar amount.

Assumptions of ANCOVA: Independence of the covariate and treatment effectImportant to ensure that:

• the covariate explains previously unexplained variance in the DV—not variance explained by the
treatment.

• the effect that the covariate has on the outcome should be the same for all of the experimental
groups.

– Randomizing participants to experimental groups or matching experimental groups on


the covariate can help.

– Important to test this before running the ANCOVA.

Assumptions of ANCOVA: Homogeneity of Regression Slopes

What ANCOVA actually does is to create a single regression equation, ignoring groups, to predict
dependent variable (Y) scores using the covariate(s) (X 1, X2, X3). Then, in essence, an ANOVA, not
ignoring groups, is performed using the residualized scores of each person computed by subtracting
each person’s predicted score based on the single equation, from the participants’ actual scores.

These computations are legitimate if the regression equations that predict the Y scores,
computed separately in the ANOVA groups, have parallel slopes. This is the homogeneity of regression
assumption, which requires that the “b” weight applied to the covariates are reasonably equal across
each group.

Testing for Equal Variance:


Different Approaches

• Levene’s Test

– If Levene’s test is significant, a more stringent alpha can be used (.01) or the variable can
be dropped from the analysis.

– This is a very conservative test and is not necessarily the best way to judge whether
variances are unequal enough to cause problems.

– Alternative Procedure: Highest and lowest variances

– Obtain standard deviation values for each of the groups; square these values to obtain
the variances.

– Take the largest value and divide it by the smallest value. If the resulting value is less
than 2, then we shouldn’t worry too much. If it is larger than 2, then we do (see next
slide).

• Calculating Effect Size for ANCOVA

• Eta squared (h2 ) is essentially r2. Derived by dividing SSM by SST. Assesses the proportion of
total variance explained by a variable.

• Partial eta squared (partial h2 ) assesses the proportion of variance that a variable explains that
is not explained by other variables in the analysis.

• Partial h2 = SSEffect

• SSEffect + SSResidual

You might also like