Graziano & Raulin
Research Methods (9th edition)
Manipulation Check
Score Data
A manipulation check involves verifying that the independent variable manipulation had the effect that the researcher intended. If the research is not experimental in nature, there will be no manipulation check. Therefore, click the PROCEED button below.
Most often, manipulation checks will involve score data, because it is common to use the more precise measures that produce score data at the experimental level. If your manipulation check produces score data, you will want to compute a mean and a standard deviation for each group and then compare the groups to see if the manipulation produced a group difference on your manipulation check variable. Links to the computational procedures for each of these statistics are given below.
Measures of Central Tendency
The most appropriate measure of central tendency for score data is the mean, although both the median and mode could also be used. [Use the back arrow key on the browser to return to the flowchart after learning how to compute these measures of central tendency.]
| Mean Computed Manually |
| Mean Computed with SPSS |
| Median Computed Manually |
| Median Computed with SPSS |
| Mode Computed Manually |
| Mode Computed With SPSS |
Measures of Variability
The most appropriate measures of variability for score data are the variance and standard deviation. The standard deviation is the square root of the variance. [Use the back arrow key on the browser to return to the flowchart after learning how to compute these measures of variability.]
Inferential Statistics
The manipulation check is designed to see if the manipulation had the effect that you intended, and if it did have that effect, you would expect the the groups or conditions would differ on the manipulation check variable. The approach that you use to do this analysis will depend on (1) how many groups you have and (2) whether the groups are correlated (as in within-subjects and matched subjects designs).
Comparing Two Independent Groups
Independent groups are formed when different people appear in the groups and the groups are selected independently of each other--that is, there is no matching of participants in the groups. If there are two groups, you have the option of using either a t-test or an ANOVA.
| Computing a t-test Manually |
| Computing a One-Way ANOVA Manually |
| Computing a t-test with SPSS |
| Computing a One-Way ANOVA with SPSS |
Comparing Two Correlated Groups
When the two groups being compared are correlated, the appropriate t-test is the correlated t-test (sometimes called a direct difference t-test) and the appropriate ANOVA is called the repeated-measures ANOVA.
| Computing a Correlated t-test Manually |
| Computing a Repeated-Measures ANOVA Manually |
| Computing a Correlated t-test with SPSS |
| Computing a Repeated-Measures ANOVA with SPSS |
Comparing More than Two Independent Groups
When there are more than two independent groups, the option of a t-test is eliminated and an ANOVA procedure must be used.
Comparing More than Two Correlated Groups
When there are more than two correlated groups, the option of a t-test is eliminated and a Repeated Measures ANOVA procedure must be used.
Are there other manipulation-check variables
that should be evaluated?
If so, click the RETURN button.
Otherwise, click the PROCEED button.