Levene’s Test of Homogeneity of Variance in SPSS (11-3)
HTML-код
- Опубликовано: 4 янв 2018
- One important assumption about the Independent-Samples t Test is that the variances in the sample groups are approximately equal. We assume that the samples have “homogeneity of variance.” Levene’s Test for Equality of Variances is a test of whether the variances of the two samples/groups are, in fact, approximately equal. You will learn how to conduct Levene’s test in SPSS both within an independent samples t test and using the Explore command.
This video teaches the following concepts and techniques:
Levene’s Test for Equality of Variances
IBM SPSS Statistics software
Link to a Google Drive folder with all of the files that I use in the videos including the Effect Size Calculator for t Tests and datasets. As I add new files, they will appear here, as well.
drive.google.com/drive/folder... 
Хобби
I could not survived my class without your RUclips lectures. Very comprehensive and well explained in the terms we can all understand. THANK YOU SO MUCH!!
Excellent! Glad that it was helpful. Thank you for the comment
Appreciate your explanation so much! I compared your pedagogical approach to several other videos and yours is by far the most effective! Thank you.
Glad it was helpful! Thank you for your kind words
Great demonstration! Thank you! Very clear!
Thank you so much! This was way clearer than my prof explained.
thank you for explaining this step to step. my professor just jumped into levene's without context. we also use spss rather than anova, so the tech instructions were also appreciated!
Glad it was helpful!
Very helpful. Amazing video.
Thank you very much
Very helpful and clear explanation. Thank you.
Thank you very much!!! Waiting for more videos. Very clear!
Thanks a lot for making this. I was struggling to find a good example with data and found this video. Much appreciated.
Glad it was helpful!
Extremely clear. Thank you so much
Explanation was so clear and concise wow. Thank you so much!
Glad you enjoyed it!
So clearly explained. Thanks a lot
This was SO CLEAR, thank you for speaking in layman's terms. I am new to statistics and also a graduate student. This was so helpful. Can you do a video like this but on ANOVA?
Thanks for the comment...glad that you liked it. Here is a playlist on ANOVA: ruclips.net/p/PLVI_iGT5ZuRk55pvafg5QDQtQrxs7IHx8
i know Im asking the wrong place but does any of you know of a tool to get back into an instagram account??
I was dumb forgot the login password. I would appreciate any tricks you can give me.
@Kyng Joe instablaster ;)
@Alden Aldo thanks for your reply. I found the site on google and I'm waiting for the hacking stuff atm.
Seems to take a while so I will reply here later with my results.
@Alden Aldo It worked and I finally got access to my account again. I'm so happy:D
Thanks so much you saved my account !
Thanks for the excellent and very clear lecture. I have a question; can I use the same protocol on SPSS while I want to check the homogeneity of variance for three groups (for checking the assumption of using one-way ANOVA) and not two groups?
thanks for that sound effect in the starting to make it less boring:)
Thank you, you're the best!
Thank you for your clear explanation and a good watching experience :)
Glad you enjoyed it!
Very clear. Thank you.
thank you so much... its a big help.
How do I use age as a grouping variable if there is more than two ages though?
I know that homoscedasticity assumption should be fulfilled as well in one way ANOVA. if the assumption was violated after looking at levene's sig level and we ran Robust Tests of Equality of Means table using both welch and Brown-Forsythe and both their sig level is less than 0.05 does that mean that the assumption is still violated and can we interpret the anova table or we should stop (even though the sig in the anova table is less than 0.05) since the assumption is not met.
saved my life! Thankyou.
Wonderful...glad to hear it. thanks for watching.
Can we use F test?
F test to find difference between the two population variance?
.05 shows up here as if were the threshold for significane but shouldn't it be .025 bc alpha is assumed to be .05, TWO-TAILED (not one-tailed)? Thank you!
Hi.. tk for your video
I just wondering if i have 2 VD and 1 VI doing homogenity test for regression
Do i put just VI to levene test or i can put all my variabel testing levene ? Tk
Thank you so much
Awesome explanation. I particularly appreciate the breakdown of the word itself, Homogeneity, to emphasize the meaning of the concept. Do you have a video for explaining Bonferroni's test?
Video on Bonferroni would be a good addition....I think I mention it with post hoc tests and ANOVA, maybe? Bottom line: divide alpha (typically p = .05) by the number of tests to get Bonferroni alpha. It works okay with 3-5 groups, but you start getting more Type II errors if you have more than that. I'd recommend Tukey's least significant difference (LSD) for post hoc if you have the option.
Good sir !
Does paired sample t test also need an homogeinity test?
Thank you.
Thanks, sir. May God bless you. You're a genius
Thank you so much for the comment and I am glad that the video was helpful.
I need to know if the variances are not equal then what should be done??Is it necessary to go for non -parametric test in that case or simple the hypothesis can be accepted or rejected based on equal variances not assumed p value.
If the variances are not equal, you simply use/interpret the Welch's t test (second line of the SPSS output "Equal variances not assumed"). It is not labeled as Welch's, but that is what the second line is. Honestly, it is a good idea to use Welch's t test regardless, instead of Student's t test, and Welch's is the default test for R.
@@ResearchByDesign yes i know that we should report the second line, but in second line, there aren't F and Sig content. so I should write F(?,?)=? , P=...
Please bring an example for me thank you so much
Dear Research By Design,
Thank you very much for making this video and it is really useful to students and beginners like me.
In the 4:58 mark of the video, you indicated and I quote, "As long as N>30 and n1 = n2, t-test is robust to violations of homogeneity variance".
1. If n1 >30 & n2 > 30 and n1 = n2 (for example n1 = n2 are both equal to 31 samples) does it mean that we NO longer need to do a Levene's Test? (i.e. that Levene's Test will be conducted only if both n1 and n2 sample sizes are less than 30?)
2. What is the best reference that I can possibly use that specifically states this "rule of thumb" (i.e. book chapter or journal publication that I can directly quote the rule of thumb)?
I plan to use Levene's Test to do a non-response bias test for early versus late respondents in my cross sectional survey of sample size 226. My samples are managers/executives of Logistics Firms (i.e. 1 manager representative per firm).
Hoping that to get clarifications on this regard.
Yours sincerely,
Haerold
I would still do the Levene's test and report the results. You always have the option of doing Welch's t-test (although not stated, this is the t test on the bottom row "equal variances not assumed" in SPSS). Be careful about interpreting findings is the p is very close to .05, however. Good luck
Good explanation sir
Thanks and welcome
EXCELLENT VIDEO!!!!!!!!!!
Thank you for the comment...hope that you keep watching!
@@ResearchByDesign New subscriber
Please, any idea why my Levene's homogeneity of variance test is showing "based on mean" median etc.
Is there anyway I can remove the mean. median?
Have you found out anything on this? It is confusing me
Yes I have, it depends on the version of SPSS you are using, but either way, it is the same test, just with different reference points. If your data is normally distributed, use the mean. If not, use the median.
Thank you Sir
Most welcome
Thanks a lottttttttt
Can you please explain what you meant by the degrees of freedom between is always 1?
Also what is 1, 198 ?
Sorry, am pretty new to these
Thank you
thanks for the question. The F(1, 198) refers to the degrees of freedom for the Levene's test (which is an F test, like an ANOVA). The first df (between) is always 1 because the t test can only be used with two groups (df = n-1). So the first df will always be 1 and the second df will always be n1 - 1 + n2 - 1, or n1 + n2 -2 (imagine that the 1 in the n1 was a subscript). good luck
what to do when the result shows a WARNINGAll absolute deviations are constant within each cell. Levene F statistics cannot be computed. what does that mean?
With any error message, try copying the text and googling it in quotes. I don't know the interpretation on this one except to say that I sounds like all of your variable scores are the same as the mean, i.e. a constant. Go back to your variables and examine each with descriptive statistics...see if they are all as you expect and that you have some variability in them. Good luck.
Hello! May ask whether it is possible to conduct levenes test for models with multiple factors or multiple independent variables??
yes, it is possible to use Levene's with more complex models, but not with the SPSS t-test version of the Levene's test. Levene's is just a F-test of population variance. Excel has a tool to compare variances, but does not call it Levene's. Look for information about variance tests and you will probably come up with what you need. Good luck
Good job.
Thank you! Cheers!
sorry but you didn't explain that how we should report the result of Leven Test, if the the Variances were different (Sig was less than 0.05)?
Good question...typically, if the Levene test is significant (i.e., p < .05, variances are not equal), then you do a Welch t test instead of Student's t test. Often, people use p < .01 or even p < .001
And what happend if you have significant in Levene's test and two factors?
i have same question
so if N>30 and t n1 = n2 I don't have to do a levene test?
If n > 30 for each group and n1 = n2, then the t test is robust to violations of homogeneity of variance. That said, if you interpret the "Equal variances not assumed line" (which is actually a Welch's test), you are always better off. In the case you describe, skipping the Levene's test would not matter. Welch's should agree with Student's t test, anyway.
@@ResearchByDesign thanks
you are adorable Dr. Daniel
Awww...so kind of you to say!
Can you please come teach my Stats class? I can actually understand what you are saying. Thank you
what if you degrees of freedom is 2
If your degrees of freedom is 2, then you have 3 (or maybe 4) participants and you have a very small sample size. You would want to add participants to increase the power of the test.
Research By Design it’s for my disertation my supervisor told us that that unfortunatly we lost some of the data and the now it’s 44 butt still accpedtable .... when I write would I still write F(2, ect....
Subtitles are coming over the SPSS results.
The only way to fix that is to turn off the subtitles while you examine the output. If you need to, you can turn them back on when you see what the output look like. Thanks for commenting!
@@ResearchByDesign I have few questions regarding study design and which test would fit for that...can I ask ?
@@ResearchByDesign or do you have any mail id.. through which I can contact you ?
cure my depression