Hi, 

I have two different questions and would thank for anyone who can try help with them:

1. If I used intervention program to facilitate some skills. I measured these skills before and after the intervention.  My assumption is that there would be improvement in the intervention group but not in the control, that it will be in all the different measures I used and that one of the measures will explain the variance in the scores of the rest of the measures more than they would explain the variance in its scores. So actually I have two times, two groups and few measures. Can I solve this with regression? what kind? repeated measures ANOVA for multiple dependent variable sound less appealing to me.

2. Theoretical question: When we use regression, we actually build a model that assume based on the size, means and standard deviation of the sample we used how should the scores of the independent variable be distributed relative to the dependent variable/s if the independent variables perfectly predict the variance in the scores of the dependent variables. Then it take the observations in the sample and compare them to the model, by that conclude how much variance is predicted by the independent variables.

My question is whether there is analysis where the regression randomly take part of the sample (assuming that it's big enough and that both the sample that was selected and this that wasn't have similar distribution as the total sample) build a model for it and compare it first with the observations randomly used to build the model and then with the  other half of the observations and based on that find how good predictors the independent variables are?

More Gilad Sabo's questions See All
Similar questions and discussions