You need to estimate a value for the group that doesn't have that predictor variable (probably easier not using the word covariate, sometimes this is over-interpreted). The default might be the mean for the group that has this. This will affect other aspects of the model. Why do you want to include this predictor variable?

Assuming I understand what you mean the covariate should be applied to a single model, that would be post= a*pre + b*cov + intercept +error +c* group. See the following link:

https://www.stat.ncsu.edu/people/davidian/courses/st790/notes.html

Best wishes, David Booth

Hello Adam

1. I think your question is not explained properly.

2. The Covariates are characteristics (excluding the actual treatment) of the participants in an experiment.

http://www.statisticshowto.com/covariate/

3. How you will measure this random variable?

4. I think the suitable test is paired-t test.

• The t-test is used to assess the statistical significance of the difference between the means of two sets of scores. That is, it helps to address the common research question of whether the average (i.e. mean) score for one set of scores differs significantly from the average score for another set of scores.
• Because in research we invariably deal with samples of people drawn from the potential population, we need to estimate whether any difference we obtain between two sets of scores is statistically significant. That is, is the obtained difference between the two means so very different from a zero difference that it is unlikely that the samples come from the same population?
• Statistical significance is assessed by using the variability in the available data to assess how likely various differences between the two means would be if there was no true difference between the two samples. The null hypothesis suggests that there is no difference between the sample means. Differences between sample means that would be very uncommon by chance if the null hypothesis were true are said to be statistically significant. They lead us to prefer the hypothesis that there is truly a difference between the two means that is unlikely to be a chance effect due to sampling.
• There are two versions of the t-test. One is used when the two sets of scores to be compared come from a single set or sample of people or when the correlation coefficient between the two sets of scores is high. This is known as the related or correlated t-test. The other version of the t-test is used when the two different sets of scores come from different groups of participants. Refer to the next chapter if the two means you wish to compare come from distinct groups of participants.
• If you have used a matching procedure to make pairs of people similar on some other characteristics then you would also use the related t-test in this present chapter – especially if the two sets of scores correlate significantly.
• Data entry for related and unrelated variables is very different in SPSS so take care to plan your analysis before entering your data in order to avoid problems and unnecessary work.
• If you have more than two sets of scores to compare then I advise to develop related analysis of variance.
• The t-test described in this chapter is known as the related t-test. Basically this means that the two sets of scores are obtained from a single sample of participants. There should be a correlation between the scores on the two measures since otherwise the two sets of scores are not related.
• The related t-test works at its optimum if the distribution of the differences between the two sets of scores is approximately bell-shaped (that is, if there is a normal distribution). If the distribution is very different from a bell shape, then one might consider using a related non-parametric statistic such as the Wilcoxon matched-pairs test.

Regards,

Zuhair