I have two unequal samples from two different population, and I want to compare their means, which statistical technique should I use here? I have read Paired sample t test can only be used if we have sample from the same population taken at different times. Please help.
Independent Samples t-test (assuming data meets the assumptions for a parametric test) or Mann-Whitney U if not.
Hi,
The paired t-test is used when you are comparing the variable of interest in a controlled group (dependent samples). For example, a group of students are given a test before and after a course, and you and to check if the course was effective, i.e. the average score has increased after the course.
In case of independent sample, or two different gropus, you can still use t-test to compare a variable of interest. In this case, you use independent t-test and it should work even for samples of different sizes. The requirements for the use of t-test are:
- Samples obtained independently from populations with normal distribution;
- The variable must be continous;
- The distribution of both populations must be approximately equal;
- The samples should be big enough (in general > 30), with approximatelly normal distribution.
If one or more of these conditions are violated, you may use non-parametric tests. For instance, the non-parametric options are: Wilcoxon (paired/dependent t-test) and Mann-Whitney (independent t-test) tests.
Regards.
Just to clarify something implied in the previous answers, the Wilcoxon and Mann-Whitney tests do not usually compare means.
Dear Ashfaque
This table will help you. As dr. Salvatore , dr. Glen and dr. Ricardo : It depend on Normality Distribution . If the number of both samples are more than 30 or 60 according to the references.
Hi Ashfaque!
Check out the continous data comparison on this page
https://www.graphpad.com/quickcalcs/
John kåre
More specifically:
https://www.graphpad.com/quickcalcs/ttest1/?Format=SD
Jk
Thank you very much to all of you. It helped me to understand the concept. Thank for your support.
...and the variances need not be equal (or approximately equal) to use the independent t-test...but you do need to first check that with an F-test and if the variances are significantly different, make sure you use the independent t-test that specifies that.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques