I am looking at the effect on an intervention program for high school students using a convenience sample. I have standardized assessment scores from fall and spring for students. Some were offered a supplemental intervention between the two test administration periods.
Approximately 140 students did NOT participate in the intervention; 64 did.
I have two key null hypotheses:
1. There is no difference in improvement rates between students who participate in the intervention program and those who do not.
2. Among students who participated in the intervention (treatment group), there is no difference in improvement rates between students who are socioeconomically disadvantaged and those who are not. [Note that I have data to separate the scores of the treatment group into these two categories].
A. I need assistance in determining the best statistical test to use for each hypothesis and which specific sets of data to compare on the test.
B. Also, should I run the tests using the actual pre- and post-test scores for each group, or should I run these on the rate of improvement (a positive, negative, or "zero" number).
I appreciate the assistance and detailed help. I believe I am overthinking this!
Thank you.
I'll say it again: you cannot use a paired-comparison t-Test on the control group versus the intervention group.
The simplest approach would be to subtract the pre-test score from the post-test score to calculate a difference score. You could then to a t-Test to compare the amount of gain in the two groups. In doing so, be sure to examine the assumption of equal variances, which is what really matters -- not the difference in sample size.
Thanks for the question
You can use Student-t test to compare the means of the scores of different groups or the same group with different criteria as you mentioned .
The idea is to find if there is a statistically significant difference ( p-value < 0.05) which will reject the null hypothesis.
The tests can be found in excel, spss and other statistics softwares.
Best wishes
Paired t-test would be suitable for this analysis. Please make sure that the following assumptions are met:
Paired comparison t-Tests would work within the separate groups (to tell you if there was significant change in either of them), but they would be appropriate for comparing the control and intervention groups.
Use paired together test for pre-post comparison and independent t test for comparing treatment and control group after checking for normality and homogeneity of variance. If not normally distributed for pre-post comparison use Wilcox test and mann Whitney u test for treatment and control group means
Normal distribution is not nearly as important as the assumption of equal variances, which should definitely be tested.
I'll say it again: you cannot use a paired-comparison t-Test on the control group versus the intervention group.
Igual you have two groups with at least two measures they vestido approach is test time by treatment interaction, in spss this can be done using GEE, mixed effect models or even ANOVA for repeated measures. They problem of using paired t test is that you won't end up answering your main question (slope difference between groups), and also counfounders can't be added
- Badges
- Science topic
- Similar topics
- Education