Two new courses are now available on my website if you want a more in-depth look at conducting a one-way MANOVA, and conducting a two-way ANOVA at TheDoctoralJourney.com/store/
Hello Prof. Can i do one way Anova with individual Likert items (Measured on 5 point scale), which are part of a formative construct in my study? I am confused whether i can go with ANOVA, as Likert items are ordinal in nature. Request you to please clarify and provide an appropriate reference.. Regards, Karan
You mention in your tutorial that for all ANOVAs, the dependent variable should be continuous. What do you mean by that? My dependent variables are all nominal groups.
+Shital Desai If your dependent variables are nominal then you can not use an ANOVA. You need to consider a nonparametic analysis. For example, a chi-square can be used when variables are nominal.
+The Doctoral Journey Greetings Dr Amanda J. Rockinson-Szapkiw! Thank you for your uploads, all of which are well-paced and highly instructional, especially for non-statisticians, such as myself. I'd like to ask a "naive" question, if I may, please, with respect to the above comment regarding a fundamental assumption of MANOVA, using a specific example (and please pardon the non optimal expression of the problem). For a repeated measures interventional ("educational intervention") experiment, on a SINGLE population (one group, ie. the same population, measured pre vs.post), where EIGHT DEPENDENT variables, at TWO LEVELS each and further divided into two sublevels (with pre post level for each; total of 36) as well as an additional one Level with no sublevel (total 8) (grand total of 36+8 = 44 ), I suppose it could be analysed using Manova IFF a high level order of measurement (interval or ratio) is used. Can one still attempt to use MANOVA for ORDINAL type measures as concerns ALL the variables and as obtained from a LIKERT scale (same 5-item scaling, but DIFFERENT responses, depending on the sublevel), given that recent research seems to be in favour of allowing the use of ANOVA in such cases (please see the reference below). If NOT, WHAT then can be used as an "equivalent" to MANOVA, for a "one-shot" analysis of such ORDINAL variables (there is a technique known as ODA, but I'd like an equivalent in "classical" statistics, if possible). Thank you very much and "bonne continuation"! G. Electronic Journal of Business Research Methods Volume 12, Issue 1, 2014, Pages 16-26 The impact of non-equidistance on Anova and alternative methods (Article) Lantz, B. Chalmers University of Technology, Gothenburg, Sweden Abstract The normality assumption behind ANOVA and other parametric methods implies not only mound shape, symmetry, and zero excess kurtosis, but also that data are equidistant. This paper uses a simulation approach to explore the impact of non-equidistance on the performance of statistical methods commonly used to compare locations across several groups. These include the one-way ANOVA and its robust alternatives, the Brown-Forsythe test, and the Welch test. We show that non-equidistance does affect these methods with respect to both significance level and power, but the impact differs between the methods. In general, the ANOVA is less sensitive to non-equidistance than the other two methods are and should therefore be the primary choice when analyzing potentially non-equidistant data.
+Saint Grégoire d'Arménie Monastère An assumption of a MANOVA is that the dependent variables (DVs) are significantly associated as demonstrated by significant correlation analysis results. If you have significantly associated DVs, which you examined multiple times across one group, a within-groups MANOVA may be appropriate. There are many factors that contribute to choosing an analysis, so based on limited study information, I can not make a specific recommendation.
+Amanda Rockinson-Szapkiw As always, you drive to the "heart" of the problem, and I wish to thank you for your response. I realise that some "gunshot" analysis (MANOVA or some MultiVariable Repeat Measure Analysis) of variables that theoretically SHOULD be dependent/associated (because they deal with similar items or ideas), is not to be attempted unless this ASSOCIATION is first verified by other means. An interesting topic in researchgate, where I found the above citation, revolves on the issue of some NON PARAMETRIC equivalent to MANOVA analysis, for which a clear answer has yet to be given. Perhaps if one was to send you a data file (let's say in SPSS) you could "quickly" provide some answers for the minimal fee (4 hours' worth) as announced on your website? Of course, no one can predict the exact level of difficulty, but with your experience, given a simplified version of the problem N=60, 5 point Likert scales throughout, 8 items (named P1 to P8) with 2 levels each (named L1, L2), and 2 sublevels (named PRE and POST), you could probably suggest a quick analysis to "prove" what otherwise looks like a flagrant POST>PRE result in the data... Once again, thank you for kindly sharing your competence.
I was having a hard time trying to decide on which test to use. Your video and examples were a great help. Thank you!
Two new courses are now available on my website if you want a more in-depth look at conducting a one-way MANOVA, and conducting a two-way ANOVA at TheDoctoralJourney.com/store/
bless your entire soul for this video it's so good
Thank you for taking time to explain these..
+nilardabbie You are welcome.
Thank god a pleasing voice. The more I learn the worse it gets
You did an excellent job of explaining these processes.
Thank you so much for this. It has clarified me on which test to use.
This tutorial is very clear and easy to follow. Thanks!
Thank you. Very nicely explained
Naren Selva You're welcome.
Amazing explanation and comparison :)
Thank you! This video is very helpful.
+Judi Barnes You are welcome.
thank you sir for explaining about this issue.
thank you~! needed these for exam
MANOVA at 9:20
the tutorial was very helpful. thanks
This is a fantastic lesson, thank you.
From 10:29 to 10:32 you say "he or she needs to consider a one-way ANOVA." I think you meant to say "a one-way MANOVA."
thanks ... helped a lot
Pretty good! A little unclear towards the end.
neeeoowww this is very helpful :)
Thanks ,but can you give an example about MANOVA how we can calculate the F
nice tutorial
Hello Prof. Can i do one way Anova with individual Likert items (Measured on 5 point scale), which are part of a formative construct in my study? I am confused whether i can go with ANOVA, as Likert items are ordinal in nature. Request you to please clarify and provide an appropriate reference.. Regards, Karan
Very helpful, thanks :)
You mention in your tutorial that for all ANOVAs, the dependent variable should be continuous. What do you mean by that? My dependent variables are all nominal groups.
+Shital Desai If your dependent variables are nominal then you can not use an ANOVA. You need to consider a nonparametic analysis. For example, a chi-square can be used when variables are nominal.
+The Doctoral Journey
Greetings Dr Amanda J. Rockinson-Szapkiw! Thank you for your uploads, all of which are well-paced and highly instructional, especially for non-statisticians, such as myself.
I'd like to ask a "naive" question, if I may, please, with respect to the above comment regarding a fundamental assumption of MANOVA, using a specific example (and please pardon the non optimal expression of the problem).
For a repeated measures interventional ("educational intervention") experiment,
on a SINGLE population (one group, ie. the same population, measured pre vs.post), where EIGHT DEPENDENT variables,
at TWO LEVELS each and further divided into two sublevels (with pre post level for each;
total of 36)
as well as an additional one Level with no sublevel (total 8)
(grand total of
36+8 = 44 ),
I suppose it could be analysed using Manova
IFF a high level order of measurement (interval or ratio) is used.
Can one still attempt to use MANOVA
for ORDINAL type measures as concerns ALL the variables
and as obtained from a
LIKERT scale
(same 5-item scaling, but DIFFERENT responses, depending on
the sublevel),
given that recent research seems to be in favour of allowing the
use of ANOVA in such cases (please see the reference below).
If NOT, WHAT then can be used as an "equivalent" to MANOVA, for a "one-shot" analysis of such ORDINAL variables
(there is a technique known as ODA, but I'd
like an equivalent in "classical" statistics, if possible).
Thank you very much and "bonne continuation"!
G.
Electronic Journal of Business Research Methods Volume 12, Issue
1, 2014, Pages 16-26 The impact of non-equidistance on Anova and alternative methods (Article)
Lantz, B. Chalmers University of Technology,
Gothenburg, Sweden Abstract The normality assumption behind ANOVA and other
parametric methods implies not only mound shape, symmetry, and zero excess
kurtosis, but also that data are equidistant. This paper uses a simulation
approach to explore the impact of non-equidistance on the performance of
statistical methods commonly used to compare locations across several groups.
These include the one-way ANOVA and its robust alternatives, the Brown-Forsythe
test, and the Welch test. We show that non-equidistance does affect these
methods with respect to both significance level and power, but the impact
differs between the methods. In general, the ANOVA is less sensitive to
non-equidistance than the other two methods are and should therefore be the
primary choice when analyzing potentially non-equidistant data.
+Saint Grégoire d'Arménie Monastère An assumption of a MANOVA is that the dependent variables (DVs) are significantly associated as demonstrated by significant correlation analysis results. If you have significantly associated DVs, which you examined multiple times across one group, a within-groups MANOVA may be appropriate. There are many factors that contribute to choosing an analysis, so based on limited study information, I can not make a specific recommendation.
+Amanda Rockinson-Szapkiw As always, you drive to the "heart" of the problem, and I wish to thank you for your response.
I realise that some "gunshot" analysis (MANOVA or some MultiVariable Repeat Measure Analysis) of variables that theoretically SHOULD be dependent/associated (because they deal with similar items or ideas), is not to be attempted unless this ASSOCIATION is first verified by other means.
An interesting topic in researchgate, where I found the above citation, revolves on the issue of some NON PARAMETRIC equivalent to MANOVA analysis, for which a clear answer has yet to be given.
Perhaps if one was to send you a data file (let's say in SPSS) you could "quickly" provide some answers for the minimal fee (4 hours' worth) as announced on your website? Of course, no one can predict the exact level of difficulty, but with your experience, given a simplified version of the problem
N=60,
5 point Likert scales throughout,
8 items (named P1 to P8) with 2 levels each (named L1, L2), and 2 sublevels (named PRE and POST),
you could probably suggest a quick analysis to "prove" what otherwise looks like a flagrant POST>PRE result in the data...
Once again, thank you for kindly sharing your competence.
Who is Neo?
I have the education level as categorical independent variable and the scores in 10 tests as dependent variables, should I conduct MANOVA???