comparison for 2 independent groups , t test or ANOVA

shesaishi

New Member
What test should i use for a comparison between two independent groups of N=30.
Also if a questionnaire is made with only 2 answers like yes or no per question , should i use U Mann- Whitney or a parametric test ?

shesaishi

New Member

I know this is easy, but i am a total newbie and i really need this . So if someone knows please answer me

rogojel

TS Contributor
hi,
if I remember correctly, the t-test and ANOVA are equivalent in the case of two groups but just as a matter of style I would use the t test, for continuous data. if you have two groups of binary data (yes, noj the two sample P test shouöd be the one to use.

Great picture btw!

shesaishi

New Member
What is this two sample p test and where do i find it in spss?

rogojel

TS Contributor
What is this two sample p test and where do i find it in spss?
hi,
ur welcome! the two sample P test compares the percentages of events (yes answers for example) in the two groups. Unfortunately I do not know spss so I cant answer this one but I guess the help will. Trybsearching for two sample proportion test as well.
regards

Last edited:

shesaishi

New Member
Thank you again for your reply , but i think i explained wrong . so i have these 2 groups of N=30 each and i want to compare their means. i had like 4 questionnaires for them , for 3 questionnaires i measured the means with t test as you said , but i have this other questionnaire with only 2 answers :yes=2 points and no=0 points. What test should i use to compare the means of the scores i obtained for each group? i thought that since it only has 2 kinds of scores that i should use a nonparametric test because i can`t get that normal distribution.

rogojel

TS Contributor
hi,
I would say this is still binary data regardless of whether you code it "yes" and " no" or 2 and 0 . As such, calculating averages is strictly speaking meaningless. E.g. if you code differently, like yes=10 no=0 you would get probably very different results, and your test must be independent of the way sou code.So, i would say, you have to treat those as discrete, binary data and use the proportions test.

regards
rogojel

shesaishi

New Member
Ok Rogojel , i will try that .
Thank you very much, you saved me!