Is the KS statistic valid in case of imbalanced data i.e. one class is overwhelmingly prevalent as compared to the other class.
Yes, this statistic doesn't require equal sample size, equal variance, or stuff like that. See e.g. http://www.physics.csbsju.edu/stats/KS-test.html, http://stats.stackexchange.com/questions/47498/mann-whitney-u-test-and-k-s-test-with-unequal-sample-sizes.
Yes, it is effective. The only requirement is that both samples should have the continuous distributions.
Yes, this test still produces valid results whether the data is balanced or imbalanced. No limitations in this regard.
The KS method is a very reliable test. Nevertheless, it can be a little hard on data some times. The chi-squared test sets a lower goal and tends to refuse the null hypothesis less often.
I think this book will provide you with many alternatives for performing statistical tests, and with useful explanations:
2006, Kanji, Gopal K., "100 Statistical Tests", Third Edition, Sage Publications.
Good Luck!
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis