0

I want to compare the means of two independent variables, but my problem is the large difference in their sample sizes. One of my variables consists of 560 people and the other only has 29. So can I use the t-test for independent samples in this case? If not, which test should I use?

Roger Fan
  • 223
erdal
  • 231
  • This is essentially the same as a previous question you have asked (see my previous comment). In what way do the answers on that previous question fail to address the question here? (the information that can be extracted is 'yes, though the significance and power are somewhat impacted by having both a large difference in sample size and variance') – Glen_b Sep 26 '14 at 02:25

1 Answers1

1

Absolutely. The only limiting factor is usually the sample size in the smallest group. The t-test performs quite well in reasonably small sample sizes under a variety of distributional assumptions. The issue of balance between groups is just a relic of old experimental designs and an issue of highly informative missingness when groups were balanced at baseline. In real life, we often have to compare highly imbalanced populations.

AdamO
  • 62,637
  • Just to be clear, another limiting factor is whether the assumptions are satisfied: Two independent random samples from normal populations having the same population SD – Russ Lenth Sep 25 '14 at 22:08
  • I have addressed this and no, normal distributions is NOT a necessary assumption with reasonable sample sizes. – AdamO Sep 26 '14 at 18:44