You're welcome. I would place the SPSS output in an appendix and refer to that in the main body together with a description of the result. peterstatistics.com/CrashCourse/3-TwoVarUnpair/NomNom/NomNom-2b-Post-hoc.html
Hey. Is there a way to attain significance testing across two variables in the table. Say, I am testing for happiness levels across states and across cities, I'd want to know the significance test not only within cities separately and within states separately but also between city A and state B
Sorry for the late reply. I'm on holiday at the moment. Perhaps what you mean is this: ruclips.net/video/dGva3nMEge4/видео.html However, you mention you have 'levels' which sounds like an ordinal variable. To compare two categories you could then use a Mann-Whitney U test, or if you have more a Kruskal-Wallis test. On my website, you should be able to find instructions on how to do any of these. Look at the two-variables unpaired section, and then either binary-ordinal, or nominal-ordinal, the at test and post hoc you should see an example and instructions on how to do that with SPSS. Hope this helps
Unfortunately, I'm not aware of any way to actually get the p-values. One way to estimate them would be to run this a few times with different alpha settings, and see when it no longer works.
it is possible in the current version of SPSS. But: Be careful, it's not about the COUNTS that are different between the columns, it's about the column PROPORTIONS! You can have for example 500 men being married and 500 women. There can still be a significant difference, not in the counts, but imagine that you asked 1000 men and 5000 women. So 50% (500/1000) of all men but only 10% (500/5000) of all women are married. I think this is an essential point for a correct interpretation of the z-test in this example!
THANK YOU!
Hi, many thanks for the video. How you you recommend presenting significant results from this test?
You're welcome. I would place the SPSS output in an appendix and refer to that in the main body together with a description of the result. peterstatistics.com/CrashCourse/3-TwoVarUnpair/NomNom/NomNom-2b-Post-hoc.html
Hey. Is there a way to attain significance testing across two variables in the table. Say, I am testing for happiness levels across states and across cities, I'd want to know the significance test not only within cities separately and within states separately but also between city A and state B
Sorry for the late reply. I'm on holiday at the moment. Perhaps what you mean is this: ruclips.net/video/dGva3nMEge4/видео.html
However, you mention you have 'levels' which sounds like an ordinal variable. To compare two categories you could then use a Mann-Whitney U test, or if you have more a Kruskal-Wallis test. On my website, you should be able to find instructions on how to do any of these. Look at the two-variables unpaired section, and then either binary-ordinal, or nominal-ordinal, the at test and post hoc you should see an example and instructions on how to do that with SPSS. Hope this helps
How can i extract the p values???
Unfortunately, I'm not aware of any way to actually get the p-values. One way to estimate them would be to run this a few times with different alpha settings, and see when it no longer works.
it is possible in the current version of SPSS. But: Be careful, it's not about the COUNTS that are different between the columns, it's about the column PROPORTIONS! You can have for example 500 men being married and 500 women. There can still be a significant difference, not in the counts, but imagine that you asked 1000 men and 5000 women. So 50% (500/1000) of all men but only 10% (500/5000) of all women are married. I think this is an essential point for a correct interpretation of the z-test in this example!