Dear fellow researchers,
I would very much like some assistance in the following:
I've run a study on the effects of a drug using pre and post tests on 13 subscales on a single group. As such I have easily run all the relevant paired sample t tests to find the difference between the means of the pre & post results per subscale.
However I also measured many other factors which could very well be confounders, diluting the effect size/change in score. These are mainly number of hours until completion of post-test, other drugs used in the interim between experimental drug and post-test, and number of times the experimental drug had been used by the subject in the past. The first 2 of these are the most important to me.
How then can I simply avoid running dozens of t tests all over again after having omitted the participants who have I) completed post-test too late ii) excessively used other drugs in the interim iii) have had e.g. more than 10 prior experiences of the drug in question ?
I have considered multiple linear regression - adding all the confounders as well as 'change in score / difference between pre and post test' to the model. But this doesn't seem to actually achieve the above i.e. adjust *the t statistic and significance* by accounting for the confounding variables. It only seems to provide percentage contributive values per confounder. Which I suppose is somewhat helpful, if I just report each of these after the original t test results? But I ideally want to *alter* the t statistic/significance by incorporating confounders
I also considered ANCOVA - by using the 'other drug use' as the independent variable, the 'hours to post-test completion' as covariate, and the 'change in score' as the dependent variable. However, again, this does not actually adjust for confounders - the thrust of this analysis seems to be identifying how the effect of the independent V (other drug use) on the dependent V (change in score) varies with/depends on the covariate (hours to post-test)… When the independent V in reality in my study was the single use of the experimental drug (after pretest but before post-test)
...I hope anyone can be of help here. Many thanks indeed in advance :)
Pascal
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude