Statistics – proportions test or p-test or z-test
When you find that a chi-Square test made on a crosstable has a significant result like this:
You then know that there are differences between the distributions in the cross table, in my example it means that male and female have different job category-distributions (see the cross table below):
But say that I want to test exactly which proportions differ from each other in this table. Is it any significant differences between 6.6% and 3.9%? It is probably not any significant differences between 25.9% and 24.6%.
To test this, you need the proportion test (z-test, p-test) that you can find in the cross table command (but also in Custom Tables module). You choose the sub command “Cells”:
And then you choose this:
Here is the result:
Have a look on the small letters beside the proportions for male resp female. When you have a and a, then it means that you have no differences between male and female, but when you have a and b there is differences.
Here you can see that in “Team leader” positions, male and female differ significantly (8.1% compared to 0.3%) . The same for Econom that also differ between 3.2% and 12.6% .
OBS you can only test column proportions, so plan carefully what to place in the columns.
Thanks for watching! /Gunilla