0
$\begingroup$

In my field some authors report the z statistic of a z-test in a generalised linear model and others do not. Does reporting the z statistic serve any purpose if the p value of a z-test or of a post-hoc Tukey test has already been reported?

Here is an example from R, with z values and p value:

Call:
glm(formula = Result ~ Condition, family = binomial, data = data1)

Coefficients:
        Estimate Std. Error z value Pr(>|z|)    
(Intercept)   3.6109     1.0134   3.563 0.000367 ***
Cond1        -2.1640     1.0630  -2.036 0.041776 *  
Cond2        0.7066     1.4284   0.495 0.620844    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
$\endgroup$
  • 2
    $\begingroup$ It is not clear to me what the z statistic of an F test might be. Perhaps I am just being dense. Can you edit in a snippet of text from one of the papers to illustrate? $\endgroup$ – mdewey Oct 5 '18 at 14:59
  • $\begingroup$ @mdewey Sorry for the confusion. Now corrected and expanded with an example. $\endgroup$ – robert Oct 5 '18 at 16:28
  • $\begingroup$ I see no F test though there. In a GLM tests would usually be either z for individual coefficients or likelihood ratio tests leading to a $\chi^2$. $\endgroup$ – mdewey Oct 5 '18 at 17:08
  • $\begingroup$ That's right, yes, my mistake. I already corrected it to z-test. $\endgroup$ – robert Oct 5 '18 at 17:28
  • $\begingroup$ You can derive p value from z value, but you cannot get z value from p value when the p value is for two-sided test, because you do not know z should be negative or positive. $\endgroup$ – user158565 Oct 6 '18 at 4:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.