Thanks for the positive feedback, glad the video helped. On a side note, RUclips has cancelled my partner program since I don't have enough subscribers (currently around 976, should be 1000 😞), so it would really help if you subscribe (and all your friends) 😉
Hi. Thanks for this amazing video. I performed Fisher's exact test on a 4*2 table in SPSS, and I got a significant difference (P= 0.010) and I wonder what is the post hoc test to use following that? is it the adjusted standardized residuals? and if I want to calculate the P value from each adjusted standardized residuals, how can I do that?
Thanks but it is a short video showing only how to calculate the residuals which I already know, but what I want is the next steps on how to convert or transform residuals into P values. @@stikpet
on my website there is a little Excel tool to convert z-value to p-values (sig.) at peterstatistics.com/CrashCourse/3-TwoVarUnpair/NomNom/NomNom-2b-Post-hoc.html
Good question. Would have to look into that one day. Since eventually these are all just z-tests Cohen's D might be a suitable effect size, and not too difficult to calculate. More info on it is available at www.socscistatistics.com/effectsize/default2.aspx#:~:text=For%20the%20single%20sample%20Z,by%20the%20population's%20standard%20deviation.
For the post-hoc, you said that we are doing 10 tests. But aren't we doing 5 tests? For example, for married people, isn't comparing male with female same as comparing female with male?
Interesting question. I'm not 100% sure, but from my understanding you are actually not comparing the male and female, but rather the observed vs. the expected counts in each cell. With only two columns (or with two rows) the number of tests could indeed be cut in half since the other column (or row) should then produce the same result, but for larger tables this won't be the case.
@@SyedaTasnia you probably need a column proportion test. Or 'simply' do a proportion test for each row. I don't have a video on this, but perhaps this site helps: www.statology.org/two-proportion-z-test-excel/
@@stikpet Thank you very much! I will check it out. And thank you for this video, I am using this to help with my stats analysis for my undergrad research project
Hi. Thank you for the video. Can you perhaps explain, how you decipher where the statistical significance lies after the post-hoc? If I understand correctly there is a difference only in the widowed 'category' between the number of males and females reported in this category? Sorry if this is too obvious.
Although it is sometimes criticized, usually a result with a significance of less than .05 is considered significant. As shown in the video, since we are doing multiple tests here we even become more strict and will only consider it significant if the result has a significance less than .005. Indeed only for male-widowed and female-widowed this was the case in the example. Note that it actually means we had 'significantly' less female widowed than we would have expected, and significantly more male-widowed than expected. Hope this helps.
The p-value is for a two sided test, but the inverse function is for one-sided probabilities, so divide by two. For example a two-tailed p-value of 0.05 would be the probability of a z-value of 1.96 or more, or -1.96 or more. The prob. of a z-value of -1.96 or less is 0.025 and the prob. of a z-value of 1.96 or more is 0.025, together create the 0.05. Hope this clarifies it a bit.
There less than expected if the z value in the example is less than the -2.81. This is because in the formula to calculate the z values I do observed minus expected, which becomes negative if there are less observed than expected. Hope this answers your question.
Glad it helped. On a side note, RUclips has cancelled my partner program since I don't have enough subscribers (currently around 976, should be 1000 😞), so it would really help if you subscribe (and all your friends) 😉
Glad the video was helpful. On my website I try to show sources for as much as possible (peterstatistics.com/CrashCourse/3-TwoVarUnpair/NomNom/NomNom3b.html). The idea for using adjusted residuals in this case is mentioned by Sharpe: Your chi-square test is statistically significant: Now what? Practical Assessment, Research & Evaluation, 20(8), 1-10. That article also mentions a few other possibilities.
No problem. On a side note, RUclips has cancelled my partner program since I don't have enough subscribers (currently around 639, should be 1000 😞), so it would really help if you subscribe (and all your friends) 😉
Glad it helped. On a side note, RUclips has cancelled my partner program since I don't have enough subscribers (currently around 782, should be 1000 😞), so it would really help if you subscribe (and all your friends) 😉
Thank you so much for this. Very easy to follow and apply to my own data!
Thanks for the positive feedback, glad the video helped.
On a side note, RUclips has cancelled my partner program since I don't have enough subscribers (currently around 976, should be 1000 😞), so it would really help if you subscribe (and all your friends) 😉
@@stikpet Done!
Thank you so much! Your video pretty much saved me!
Glad it helped!
Hi. Thanks for this amazing video. I performed Fisher's exact test on a 4*2 table in SPSS, and I got a significant difference (P= 0.010) and I wonder what is the post hoc test to use following that? is it the adjusted standardized residuals?
and if I want to calculate the P value from each adjusted standardized residuals, how can I do that?
since you are using SPSS, perhaps this helps: ruclips.net/video/dGva3nMEge4/видео.html
Thanks but it is a short video showing only how to calculate the residuals which I already know, but what I want is the next steps on how to convert or transform residuals into P values. @@stikpet
on my website there is a little Excel tool to convert z-value to p-values (sig.) at peterstatistics.com/CrashCourse/3-TwoVarUnpair/NomNom/NomNom-2b-Post-hoc.html
Thanks for your efforts and help, I will check it and see.@@stikpet
This video is a lifesaver - thank you. Is there any way to calculate effect size here?
Good question. Would have to look into that one day. Since eventually these are all just z-tests Cohen's D might be a suitable effect size, and not too difficult to calculate. More info on it is available at www.socscistatistics.com/effectsize/default2.aspx#:~:text=For%20the%20single%20sample%20Z,by%20the%20population's%20standard%20deviation.
@@stikpet Great - thanks.
For the post-hoc, you said that we are doing 10 tests. But aren't we doing 5 tests? For example, for married people, isn't comparing male with female same as comparing female with male?
Interesting question. I'm not 100% sure, but from my understanding you are actually not comparing the male and female, but rather the observed vs. the expected counts in each cell. With only two columns (or with two rows) the number of tests could indeed be cut in half since the other column (or row) should then produce the same result, but for larger tables this won't be the case.
@@stikpet so how would this be carried out if I was comparing male vs female? And how would it be done if was comparing martial status?
@@SyedaTasnia you probably need a column proportion test. Or 'simply' do a proportion test for each row. I don't have a video on this, but perhaps this site helps: www.statology.org/two-proportion-z-test-excel/
@@stikpet Thank you very much! I will check it out. And thank you for this video, I am using this to help with my stats analysis for my undergrad research project
Hi. Thank you for the video. Can you perhaps explain, how you decipher where the statistical significance lies after the post-hoc? If I understand correctly there is a difference only in the widowed 'category' between the number of males and females reported in this category? Sorry if this is too obvious.
Although it is sometimes criticized, usually a result with a significance of less than .05 is considered significant. As shown in the video, since we are doing multiple tests here we even become more strict and will only consider it significant if the result has a significance less than .005. Indeed only for male-widowed and female-widowed this was the case in the example. Note that it actually means we had 'significantly' less female widowed than we would have expected, and significantly more male-widowed than expected.
Hope this helps.
@@stikpet Yes, now I understand. Thank you for the immediate answer!
Why did you divide significance lvl with 2 to get the z value?
The p-value is for a two sided test, but the inverse function is for one-sided probabilities, so divide by two.
For example a two-tailed p-value of 0.05 would be the probability of a z-value of 1.96 or more, or -1.96 or more. The prob. of a z-value of -1.96 or less is 0.025 and the prob. of a z-value of 1.96 or more is 0.025, together create the 0.05.
Hope this clarifies it a bit.
Hi, why is it that there are LESS than expected if it's greater than the z-critical value?
There less than expected if the z value in the example is less than the -2.81. This is because in the formula to calculate the z values I do observed minus expected, which becomes negative if there are less observed than expected. Hope this answers your question.
Ah, I understand now. Thank you for your help!
Love this -- super helpful.
glad it helped. Good luck with your analysis
Thank you so much for this ! :)
Glad it helped.
On a side note, RUclips has cancelled my partner program since I don't have enough subscribers (currently around 976, should be 1000 😞), so it would really help if you subscribe (and all your friends) 😉
Is there a function for the adjusted residual in Excel?
Unfortunately there is not.
Hey, great video! Congrats and thanks!
Could I please have the reference for this??
Glad the video was helpful. On my website I try to show sources for as much as possible (peterstatistics.com/CrashCourse/3-TwoVarUnpair/NomNom/NomNom3b.html). The idea for using adjusted residuals in this case is mentioned by Sharpe: Your chi-square test is statistically significant: Now what? Practical Assessment, Research & Evaluation, 20(8), 1-10. That article also mentions a few other possibilities.
AMAZING, my friend!
Just solved my problem!
Thanks for the help! BIG TIME!
Cheers!
No problem.
On a side note, RUclips has cancelled my partner program since I don't have enough subscribers (currently around 639, should be 1000 😞), so it would really help if you subscribe (and all your friends) 😉
@@stikpet Done, bro! 👍
Saved my day! Thank you =)
Glad it helped.
On a side note, RUclips has cancelled my partner program since I don't have enough subscribers (currently around 782, should be 1000 😞), so it would really help if you subscribe (and all your friends) 😉
Could you speak a bit louder in the next videos, please? Thank you
thanks for the feedback. Audio is indeed a bit poor, but hope it was still helpful.