0

On one of my questionnaires, which measure learning behaviours, after PCA, I have 4 subgroups, Efficacy/Perseverance/Effort/Achievement - I have run a T-Test and I have 4 sets of data comparing my samples of Undergrads to Graduates - Is it appropriate to only use T Test to report the results. Similar with my next questionnaire...Approaches to learning - 3 subgroups, Deep/Surface/Strategic - T Test with 3 results for each. This would mean that I would so far report 7 sets of T Test results? (+ 1 questionnaire with no subgroups)(So 8 altogether)

This has been edited to give a little more information

Tim
  • 138,066
Trudy Baker
  • 1
  • 1
    Could you elaborate a little bit more, because it is not clear what you are asking..? – Tim Feb 08 '15 at 12:32
  • On one of my questionnaires, which measure learning behaviours, after PCA, I have 4 subgroups, Efficacy/Perseverance/Effort/Achievement - I have run a T-Test and I have 4 sets of data comparing my samples of Undergrads to Graduates - So should I report all 4 results. Similar with my next questionnaire...Approaches to learning - 3 subgroups, Deep/Surface/Strategic - T Test with 3 results for each. This would mean that I would so far report 7 sets of T Test results? – Trudy Baker Feb 08 '15 at 12:45
  • 2
    I thing you should edit your initial question to add more details on your data and what are the aims of your analysis. – Tim Feb 08 '15 at 12:50
  • Could you clarify what type of t-tests and for what differences you are checking? How do the t-tests relate to your hypotheses? – noumenal Feb 08 '15 at 20:44
  • Do you mean you have two independent groups, graduates and undergraduates and that you have seven dependent variables, for each of which you like to test if the two groups differ? In that case, yes, you run seven tests to compare the two groups. But how many cases in each group? What values do the seven dependents have. These things determine if a t test would apply or some other test..... We really need more info before any advice can be given. – BenP Mar 10 '24 at 08:50

1 Answers1

0

EDIT: With PCA you can report how the (rotated) variables load onto the various factors (Efficacy, Perserverance, Effort, Achievement). By doing so you justify the grouping of variables. If an instrument/questionnaire is designed to group in this way, you could confirm/disprove that using the factor loadings. If it is an established instrument I would check how it is generally reported in the literature.

Use a correction of alpha for multiple comparisons, such as the Holm-Bonferroni method. If you have a hierarchical situation, you might be better of using a nested ANOVA for group comparisons. Just check that all the assumptions of the test are fulfilled.

noumenal
  • 642
  • The question of whether to use a 'correction' for multiple comparisons is a complicated one that is strongly dependent on the context of the experiment and analysis. See here, for example: https://stats.stackexchange.com/questions/628319/p-value-correction-in-multiple-outcomes-study – Michael Lew Mar 09 '24 at 20:15