05 May 2016 13 3K Report

Hello, I am doing an analysis of around 1600 students’ (fairly large sample) data with regard to three variables (continuous data). One variable has five dimensions (all continuous data). Now suppose I want to test the difference in means of male and females on one variable (approximately normal) but the data in one group (male) is not normal but for female it is normal (approx.). So the question is whether I can assume data as normal just because the total data of male and female is normal and then apply t test or I should use some other non-parametric (eg Mann-Whitney U) test. Is the data on both the groups must be normal for t test? Please quote reference, if any.

Thanks in advance.

Similar questions and discussions