See All Boards Subscribe To Thread
Statistics Public Forums
Feedback on Resources: Test assumptions and clear wording regarding them
umakegoodcookies Aug 08, 2016 ReplyFlag
The wording regarding assumptions is unclear in places. For example in section 3.2 where it talks about normality of "the data". Which data is that? It should be clear that the normality assumption is about the population and examining the sample is one of the ways to check the plausibility of that assumption. I think that it's implied because you're talking about the generating process but it would be helpful if that were expanded a little.

The sentence near the beginning of that section, "Testing the appropriateness of the normal assumption is a key step in many data analyses," is irksome. Testing implies some sort of hypothesis test and you do a good job at avoiding them. Perhaps we can say "examining" instead of "testing"?

It would also be great if there were some discussion of other information that can be taken into account as well. For example, if you have a small sample but know from prior large sample studies that the population of data is approximately normally then you need to weight any examination you make by considering that information. In fact, the word approximate should be used substantially more liberally regarding assumptions.
David Aug 08, 2016 ReplyFlag
Indeed, this is always a tough topic. We want to be rigorous but also not overburden readers too much. In past editions, we were more strict and conservative on language, and what we learned was that many readers basically became nervous about doing anything. So, the language has been softened in the book.

It's certainly not a perfect balance, and there's always room for improvement. If you'd like to make specific suggestions, the "Types and Feedback" form on the bottom of the Textbook page ( would be the best place to do so.


Your User ID will appear with your posts.