I discuss the assumptions of both the pooled-variance and Welch (unpooled variance) t tests and confidence intervals, and their advantages and disadvantages. I illustrate some of pros and cons using the results of two simulations.
I discuss the assumptions of both the pooled-variance and Welch (unpooled variance) t tests and confidence intervals, and their advantages and disadvantages. I illustrate some of pros and cons using the results of two simulations.
Would it make sense to repeat the first table using Welch method, to give a direct comparison with the results obtained from pooled variance (i.e. how much closer to 95% is the Welch method at the extremes)?
Yes, that would definitely make sense. I did check it at the time, and (not too shockingly) Welch was pretty close to 95% in all those spots. I make comment about that in my notes/text, but I’m not sure if I state that in this video.