Harry_the_cat
Elite Member
- Joined
- Mar 16, 2016
- Messages
- 3,693
More and more I am noticing students and textbooks using the phrase "accept the null hypothesis" in statistical analysis.
I have always insisted that students use the term "fail to reject H0" rather than "accept H0". The difference appears subtle and seems to be ignored by many textbook writers (and subsequently teachers and then students) these days. To me, this is analogous to the terms used in a court of law ie "guilty" or "not guilty" - we don't say "guilty" or "innocent".
We aren't "accepting" the null hypothesis, we are only saying that the data doesn't give us enough evidence to "reject" it. That's a very different thing.
I'd like to know others thoughts on this, particularly if you have a background in statistical analysis. It has bothered me for a while.
I have always insisted that students use the term "fail to reject H0" rather than "accept H0". The difference appears subtle and seems to be ignored by many textbook writers (and subsequently teachers and then students) these days. To me, this is analogous to the terms used in a court of law ie "guilty" or "not guilty" - we don't say "guilty" or "innocent".
We aren't "accepting" the null hypothesis, we are only saying that the data doesn't give us enough evidence to "reject" it. That's a very different thing.
I'd like to know others thoughts on this, particularly if you have a background in statistical analysis. It has bothered me for a while.