The normality test for my pretest is not normal but the scores are normal for my posttest and delayed posttest. Can I still do a paired t-test?
The paired-samples t-test assumes the difference scores (between pre- and post) are normally distributed.
Was the same sample measured at 3 time points? If so, a repeated-measures ANOVA would be better than multiple t-tests.
Were all three tests the same (or parallel forms of the same test)?
And you couldn't show the post and delayed tests were "normal", you (presumably) just didn't reject that hypothesis.
Hello Shruti,
What is the specific research question you are trying to answer? That, the nature of the scores with which you are working and how data were collected is what should inform the choice of analysis.
Both Blaine and Daniel are offering germane points to your query, but if you could elaborate on your intended analyses, the scale strength of your scores, whether scores on all occasions are on the same (or equated) measures, and the data collection, then more focused recommendations could be offered.
Good luck with your work.
Thank you all so much for the responses. I am trying to find the before and after effects of an intervention that I had implemented on a set of management postgraduate students.
The three tests had the same format though the questions within them varied. I have conducted a repeated measures ANOVA test as well.
Thank you
Thank you all so much for the responses. I am trying to find the before and after effects of an intervention that I had implemented on a set of management postgraduate students.
The three tests had the same format though the questions within them varied. I have conducted a repeated measures ANOVA test as well.
Thank you
Thank you all so much for the responses. I am trying to find the before and after effects of an intervention that I had implemented on a set of management postgraduate students.
The three tests had the same format though the questions within them varied. I have conducted a repeated measures ANOVA test as well.
Thank you
Thank you all so much for the responses. I am trying to find the before and after effects of an intervention that I had implemented on a set of management postgraduate students.
The three tests had the same format though the questions within them varied. I have conducted a repeated measures ANOVA test as well.
Thank you
Check the normality for pre-post (difference) then take a call...if the number of obs >30 as per CLT you can go ahead with paired t-test.
Try both the t-test and a non-parametric test and report both results. Don't worry about the outcome. Just let the data speak.
Best!!
AN
Since the questions varied, the pre-post measure likely does not have an expected mean of 0. Even if the items were randomly chosen from a test back, give the number of items will be relatively small, it is variation among them that will be critical to consider. More details are needed about how items were chosen and the research questions, but one thing is clear that a paired t-test (or those alternatives that do not made as many distributional assumptions) is unlikely to be appropriate. Please provide details about how the items were chosen and your research questions.
Thank you all for your responses. I have received some clarity. Thank you
- Badges
- Science topic
- Similar topics
- Correction