I actually do not not the precise method for report in APA format. Just note that youre conducted the Brown-Forsythe tests., that is was significant (add levine statistics and p value), then report the sig value with the F and p vlaues. I don't beleive their is a standard protocol in APA for doing this - at least I haven't found it.
Thank you Robin! It's very helpful since my book leaves out this part. May I ask how do I interpret my result if the p values of ANOVA, Welsh and Brown are all .000? ... Coz it feels a bit unsafe as the figures are too extreme. Thx in advance!
I suspect in this example Levene's test is coming out significant because the group n's are big, not because the group SDs different enough to really matter. (This is the downside of using a test to compare variances.) This is supported by the three F's coming out pretty similar. If you can lay your hands on R's "insectsprays" data, that would be a rather more extreme example of different spreads.
Hi, Robin. Thank you for the video. Was really helpful! A textbook that I'm referring didn't explain very well what to do if Levene's test gives significant result. By the way, I have a question for you, is there any statistical analysis that we can compare between coefficient of variations (CVs), in SPSS. For example, I have 6 CVs representing 6 different variables within a group, and I want to know if there are significantly different or not to each other. I hope you can make a tutorial video for this topic, I have been looking for this kind of analysis from many sources but couldn't find the concrete answer. Thank you.
Great video. Thank you. Is there an equivalent for this in a MANOVA framework? I have a couple of DVs where Levene's is violated and I have unequal sample sizes.
Also, back now just to an ANOVA with one DV that violates Levene's: if the B-F and Welch's come back significant, can we then interpret the post-hoc analyses as normal?
I didnt understand, does that mean we should use the significant level in welch rather than the one in ANOVA? but what about the assumption, isnt it still violated since we still got sig level >0.05 in the welch and the other test? I'm confused :/
That is a great viedo, thanks a lot. Now, which degrees of freedom do you report in this case: F(2,163) or perhaps the ones you see on the Welch table F(2,68.465)?
Um...I am confused. When interpreting the Levene's test, doesn't the sig. value have to be ABOVE .05 for the results to be counted as statistically significant meaning the homogeneity of variances is not violated? Yours is below but you said it IS significant and that that assumption was not violated. :/
I jusu wacthed again - it's been a while. I said the the test for homgeneity of variance is significant because the p value is less than .05 - That means the variances are not the same so we have to look at the stricter tests talked about at 2:35.
What about if the result: sig in test of homogeneity = 0.000 ANOVA sig = 0.35 Welch sig = 0.56 Brown-Forsythe sig= 0.30 Could you help me interpret this result? Coz you said Welch is preferred if there's any difference...
So there is a sig. diff among the groups you are comparing with your ANOVA based on the Levine test. The ANOVA says p< 0.05 BUT the Welch test says p > 0.5 so you conservatively conclude there is no sig diff among your group variable. From what I can tell there is no hard and fast rule about this - If your sample sizes are relatively equally, then you could use the BF tests. See statistics.laerd.com/statistical-guides/one-way-anova-statistical-guide-3.php from SPSS.
Hi, I was hoping you could help me! At 2.59 "more rigorous test to be used when the variances are not equal', which variances are this referring to? Cheers
The variances (or spread of values) for each group you are comparing. So each group has a mean and standard deviation (variance equals SD squared) and you want the standard deviations to be roughly similar If the groups do not have roughly equally variances, then there is a stricter test that needs to be applied
So my Levene's is .011 (cant report my ANOVA) and my Welch is .197 and Brown-F is .175... What do i do? The Levene's is significant and the others are not.
So what that means is that there are no significant differences among group sin your ANOVA - The levine test says "Don't use the regular ANOVA test - use the statistics for the Brown-F and/or Welch - In your case, those statistics suggest that there are no significant difference among your means.
Explained clearly, thank you!
Hey,Robin! You were very sharp in your explanation. Thank you!
+Humberto Fioravante Ferro Thanks and happy to help
BTW - This test was conducted using SPSS - Not sure of the equivalent test for other packages like Mintab or R
Excellent information. Thank you so much!!
+Robin Kay Please cite a reference for justification of using Brown-Forsythe and Welch tests in case of failing Homogeneity of Variances.
This is from Discovering Statistics Using SPSS - Andy Fields (p. 347)
Robin Kay Thanks a lot
Thank you! Life saver
So glad I good help, even if it takes me a year to respond ;-)
Thank you very much for the explanation!
You are welcome ( a year later).
Hi Robin, how would you report Brown-Forsythe APA style once we have concluded that is non significant either? could you give an example? please
I actually do not not the precise method for report in APA format. Just note that youre conducted the Brown-Forsythe tests., that is was significant (add levine statistics and p value), then report the sig value with the F and p vlaues. I don't beleive their is a standard protocol in APA for doing this - at least I haven't found it.
Thank you Robin! It's very helpful since my book leaves out this part.
May I ask how do I interpret my result if the p values of ANOVA, Welsh and Brown are all .000? ... Coz it feels a bit unsafe as the figures are too extreme. Thx in advance!
This would mean that you would assume unequal variances
Goodness - I droppe dthe ball on this reply- sorry. I don't think you need to worry if the p value says 0.000 - it just means it is very small
How to deal with unequal sample sizes in a two way ANOVA. Please help.
I suspect in this example Levene's test is coming out significant because the group n's are big, not because the group SDs different enough to really matter. (This is the downside of using a test to compare variances.) This is supported by the three F's coming out pretty similar. If you can lay your hands on R's "insectsprays" data, that would be a rather more extreme example of different spreads.
ps: I appear to be signed into youtube as my daughter.
Hi, Robin. Thank you for the video. Was really helpful! A textbook that I'm referring didn't explain very well what to do if Levene's test gives significant result.
By the way, I have a question for you, is there any statistical analysis that we can compare between coefficient of variations (CVs), in SPSS. For example, I have 6 CVs representing 6 different variables within a group, and I want to know if there are significantly different or not to each other. I hope you can make a tutorial video for this topic, I have been looking for this kind of analysis from many sources but couldn't find the concrete answer. Thank you.
Great video. Thank you. Is there an equivalent for this in a MANOVA framework? I have a couple of DVs where Levene's is violated and I have unequal sample sizes.
Also, back now just to an ANOVA with one DV that violates Levene's: if the B-F and Welch's come back significant, can we then interpret the post-hoc analyses as normal?
Oh, I see in the link I should use a Games Howell or Dunnett's C post hoc test. Thanks again.
I didnt understand, does that mean we should use the significant level in welch rather than the one in ANOVA? but what about the assumption, isnt it still violated since we still got sig level >0.05 in the welch and the other test? I'm confused :/
That is a great viedo, thanks a lot. Now, which degrees of freedom do you report in this case: F(2,163) or perhaps the ones you see on the Welch table F(2,68.465)?
I would go with 2, 163
Um...I am confused. When interpreting the Levene's test, doesn't the sig. value have to be ABOVE .05 for the results to be counted as statistically significant meaning the homogeneity of variances is not violated? Yours is below but you said it IS significant and that that assumption was not violated. :/
I jusu wacthed again - it's been a while. I said the the test for homgeneity of variance is significant because the p value is less than .05 - That means the variances are not the same so we have to look at the stricter tests talked about at 2:35.
@@drkayotu thank you
What about if the result:
sig in test of homogeneity = 0.000
ANOVA sig = 0.35
Welch sig = 0.56
Brown-Forsythe sig= 0.30
Could you help me interpret this result?
Coz you said Welch is preferred if there's any difference...
So there is a sig. diff among the groups you are comparing with your ANOVA based on the Levine test. The ANOVA says p< 0.05 BUT the Welch test says p > 0.5 so you conservatively conclude there is no sig diff among your group variable. From what I can tell there is no hard and fast rule about this - If your sample sizes are relatively equally, then you could use the BF tests. See statistics.laerd.com/statistical-guides/one-way-anova-statistical-guide-3.php from SPSS.
When homogeneity of variance is violated, and Brown-Forsythe and Welch are significant, are the post-hoc analyses for ANOVA (i.e. Sheffe etc.) valid?
Amichai Perlman - Yes if they are significant, the post hoc tests are valid.
Which post-hoc test would you recommend?
A little late in responding - AIt depends, but the Games-Howell is popular. I like this video on that test: ruclips.net/video/oagLeAOaevk/видео.html
Hi, I was hoping you could help me! At 2.59 "more rigorous test to be used when the variances are not equal', which variances are this referring to? Cheers
The variances (or spread of values) for each group you are comparing. So each group has a mean and standard deviation (variance equals SD squared) and you want the standard deviations to be roughly similar If the groups do not have roughly equally variances, then there is a stricter test that needs to be applied
So my Levene's is .011 (cant report my ANOVA) and my Welch is .197 and Brown-F is .175... What do i do? The Levene's is significant and the others are not.
So what that means is that there are no significant differences among group sin your ANOVA - The levine test says "Don't use the regular ANOVA test - use the statistics for the Brown-F and/or Welch - In your case, those statistics suggest that there are no significant difference among your means.
So how do I report that in my paper write up? I'd ask my professor but I don't want them to think I'm statistically challenged
"Levene's test for homogeneity of variances was significant (p
Please cite any reference if you can
I used Discovering Statistics by Andy Field = Amazing text for Stats and SPSS.