1
$\begingroup$

I am trying to understand analysis in a paper which talks about statistical significance of a certain stimulus. I have used 'VALUE' in place of the specific variable that is being observed, and 'STIMULUS' to represent experimental stimulus on the experiment group that may influence the VALUE people get. The following is what it says:

We find that 89% of those who got their VALUE increased were given the STIMULUS (experiment group), compared to 64% who did not get the VALUE increased. This difference is significant (p=.000065, χ2=16.02), and suggests that STIMULUS impacts the VALUE people get.

I would like to understand how the author has calculated the significance. What kind of significance test is this?

$\endgroup$

1 Answer 1

1
$\begingroup$

It's certainly a test based on a statistic assumed to have a chi-square distribution with one degree of freedom under the null, asymptotically at least; & the description makes it clear two counted proportions are being compared. Wald, score, & likelihood-ratio tests are all possibilities; Pearson's chi-squared test for homogeneity is the most popular (& Yates' correction may or may not have been used).

Asymptotically these tests are equivalent—in this case the information given suggests sample sizes sufficiently large that any difference in the p-values got by using one or the other would be unimportant for most purposes.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.