what type of statistical analysis would you apply?

#1
I am a teacher. I've done pre and post intervention testing with a group (n=17) of pupils. I also did the same tests with another group (n=17) who had no intervention, they just had business-as-usual teaching.

Whats the first thing you would do - compare the initial test scores for the two groups using a t-test for two samples (with unknown variances)?
 

Karabiner

TS Contributor
#2
Comparing baseline scores is a good idea. With n=34, statistical power will be low, though, therefore a nonsignificant result should not be overinterpreted.

What to do depends on the particular topic of your research, the nature of the measurements taken, and on your research question.

With kind regards

K.
 
#3
Thank you Karabiner. I can offer some more info. The research question - "Does (special kind of story-telling) increase literacy skills more than business-as-usual teaching?"

The data is from a standardised test. This test is a little different from your usual psychometric, possible test scores run from 0 to 20, the standardised population average is 10. there are a number of domains within the test, I have tested the children on five of these, before and after intervention - so I have 10 data points for each child involved. So what do you think, would you start with the t-test?
 

hlsmith

Not a robit
#5
You would want make sure the student groups were comparable on covariates of interest, including scores. What was the time gap between the two tests, short enough to solely attribute changes on intervention?
 

CowboyBear

Super Moderator
#6
The standard analysis for this situation (pre and post measurements on two groups) would be a repeated measures ANOVA. Your main interest would be the interaction between time and group (i.e., does the intervention group change more than the control group?)

It's also possible to address the question with a multilevel/mixed model.
 
#7
there was a 10 week gap between pre and post intervention, and the control group was tested simultaneously. which of the various t-tests available are appropriate to test the comparability of the two groups? From reading around the topic I think it might be a 2 group comparison with unknown variances - which I can do using excel I think)