Multivariate ordinal logistic Regression looks like what you need. Try this link:

https://www.google.com/search?q=multivariate+ordinal+logistic+regression&rlz=1C1CHBF_enUS847US847&oq=multivariate+ordinal+logistic&aqs=chrome.0.0j69i57j0.25568j0j7&sourceid=chrome&ie=UTF-8

Best wishes, David E BBooth

If your dependent variables are categorical, multiple logistic regression is suitable here.

I would examine whether it is possible to combine your dependent variables into a single scale. Because there have been so many questions here on this topic, I have compiled at set of resources on it:

https://www.researchgate.net/post/How_can_one_help_people_create_simple_scales_from_Likert-scored_items

Sir, I think multiple logistic regression, discriminant analysis, or neural network analysis will help for your research.

One possibility for the outcome variables is to perform factor analysis to reduce the number of variables. Ordinal logistic regression per outcome variable may be useful. If your sample is large (hundreds) CHAID or CART could be useful. Ultimately, the methods you use are determined the goal you want to achieve.

You can use multivariate logistic which handles multiple responses simultaneously without conbining them into a single score - here is an example

Article Investigating the Macro Determinants of Self-Rated Health an...

This methodology will handle situations when not all the responses are observed as long as one is, and you make MAR assumptions which are not too demanding:

This is an extension of a multivariate multilevel model:

https://www.researchgate.net/post/Can_you_share_any_examples_of_the_use_of_multivariate_regression

See

https://stats.idre.ucla.edu/other/mult-pkg/whatstat/

If you are a relative beginner at regression analysis, then it will be very demanding to learn about advanced alternatives such as "multivariate ordinal logistic" and "multivariate logistic" regression.

This is why I made the suggestion to try forming a single, scale variable to use as the dependent variable, which you can then use in an ordinary least square regression.

David L Morgan David, What you say is of course correct. However I don't see how your suggestion helps Dr. Patel solve his current problem. He has multiple DVs not some kind of function of DVs. We should not advise him to change his research to find an easier method if we have something that will solve his problem. In this case Multivariate ordinal logistic regression would seem to do that. I am sure that he is capable of reading the material I suggested. Best, David Booth

It dépends of the phenomena you really want to study. Do you want a comprehensive description of interrelations between those variables ? If it is the case, use Principal components analysis (PCA) or even a correspondances analysis (but I will have to split the data table). It will give you a good view of the inter-relations. You can use free FactoMiner package in R environnement (using Rstudio) for instance or Excel add on such as XLStat software.

David Eugene Booth, I don't think we know enough about the dependent variables from this description to make an adequate recommendation, which is why I labelled my a remarks about forming a scale a "suggestion." Of course, one could always use multi-variate ordinal regression, but if the variables do form a scale, then that would be like using a steam shovel to empty a sandbox. Hence, my suggestion to investigate simpler solutions first.

David L Morgan Well my old algebra professor used to say many of my proofs used elephant guns to kill gnats. I always believed that my job was to solve a problem. If other interesting problems came up in the meantime they could be attacked later. I do not dispute that something like what you suggest may be possible only that it is not the most direct way to the goal IMO. Best wishes, David Booth

I see a three-factor MANOVA with 7 dependent variables at first glance. It is also possible to have one or two covariates from your IVs, in which case a MANCOVA may ensue. I don't know what the DVs are and how interrelated they might be. The assumption of multicolinearity in multivariate statistics is a big issue and might reduce the number of DVs by way of providing solution. Otherwise, do a dimension reduction with principal component analysis and get on with your work.

Maurice Ekpenyong This can't be a MANOVA because the outcome variables are ordinal and thus there is no place for a MV normal distribution to come in. Best, David Booth

Sorry Prof. Didn't look at the nature of variables but just the number. Didn't have time to edit the answer. Sincere apologies please and thanks.

SPSS- Statistica Programming Software

use logistic regression analysis for your data it is quite easy and give meaningful conclusion. you can use SPSS or STATA software Nadia H. Al-Noor

Sudip Poudel does spss perform multvariate (not multiple) ordinal regression, I would like to know.

Hello Kelvyn Jones. AFAIK, SPSS has no canned procedure to estimate a multivariate ordinal regression model. But your question reminded me of these old threads in the SPSSX-L discussion forum:

• http://spssx-discussion.1045642.n5.nabble.com/Multivariate-Analysis-in-MIXED-tp1092023.html
• http://spssx-discussion.1045642.n5.nabble.com/Multivariate-Mixed-Model-Follow-Up-td3392871.html

Obviously, the topic is multivariate linear models, not multivariate ordinal regression models. But it makes me wonder if the same approach (i.e., using a categorical variable to indicate which DV is recorded on a given row) would work. I've not tried it with PLUM or GENLIN, the SPSS commands that can be used to estimate ordinal logit models.

• https://www.ibm.com/support/knowledgecenter/en/SSLVMB_26.0.0/statistics_reference_project_ddita/spss/advanced/syn_plum.html
• https://www.ibm.com/support/knowledgecenter/en/SSLVMB_26.0.0/statistics_reference_project_ddita/spss/advanced/syn_genlin.html

If someone can point to an example of a multivariate ordinal regression model (with the data available), I might be tempted to try it using this approach.

Out of curiosity, I just did a search for multivariate ordinal regression with Stata (the other software package I use). A Statalist thread from 2014 suggests that generalized SEM (-gsem- command) is the way to go. SPSS users tend to use AMOS for SEM. I don't know if it is able to handle categorical outcome variables.

• https://www.stata.com/features/overview/generalized-sem/

HTH.

EDIT: Upon further reflection, I think one would need to use the GENLINMIXED command to try this approach in SPSS.

• https://www.ibm.com/support/knowledgecenter/en/SSLVMB_26.0.0/statistics_reference_project_ddita/spss/advanced/syn_genlinmixed.html

The best solution i will recommend is to convert the responses on the Likert Scale to continuous data. Since your independent variables are categorical, Three-way Multivariate Analysis of Variance (MANOVA) can now be employed. Alternative you could run 21 separate ANOVA using one IV to 1 DV per time. Or you can use PCA by spliting the output as suggested earlier by Olivier Las-Vergnas

Valentine Joseph Owan If likert scales could be made continuous ordinal logistic regression would never have been invented

Owan is right Likert scale response could be converted to discrete or continuous data. This is done through rank scoring and summing based on clubs or grouping. That's is in my opinion need to be done to address this problem

Book Reviews : Maurice M. Tatsuoka. Selected Topics in Advanced Statistics: An Elementary Approach. Champaign, Ill.: Institute for Personality and Ability Testing. No. 1. (Not available) No. 2. Standardized Scales: Linear and Area Transformations, 1969. Pp. iii + 15. $1.00 (paperback). No. 3. Classification Procedures: Profile Similarity, 1974. Pp. iii + 47. $2.50 (paperback). No. 4. Significance Tests: Univariate and Multivariate, 1971. Pp. iv + 63. $2.50 (paperback). No. 5. Validation Studies: The Use of Multiple Regression Equations, 1969. Pp. iii + 28. With supplement, What is Job Relevance?, 1973. Pp. iii + 16. $1.25 (paperback). No. 6. Discriminant Analysis: The Study of Group Differences, 1970. Pp. iv + 57. $2.50 (paperback). No. 7. The General Linear Model: A "New" Trend in Analysis of Variance, 1975. Pp. iii + 64. $3.00 (paperback). Complete set of six booklets. $10.25 (paperback

Walsh, J. A.

https://booksc.xyz/book/39658872/ed5c5c

Hi Patel,

It is a likert scales. You can find a new variable representing the sum of the responses and then apply the analysis of variance with three factors to see the effect of these factors and their interactions on the new variable, which represents the sum of the responses. Also, you can apply a regression analysis.

Best wishes

I believe three-way Multivariate Analysis of Variance (MANOVA) can be utilized. Structural equation modeling is a multivariate statistical analysis technique that is used to analyze structural relationships.

اعتقد يستخدم اختبار manova في تحليل مقياس ليكرت

Some may be interested in this recent paper the difficulties of simply adding up score on different multiple responses

Article Understanding the population structure of the GHQ-12: Method...

I suggest you use MANOVA

Kelvyn Jones, does the GHQ-12 use different response formats for its sub-scales? If so, then of course adding up the sum of the sub-scales is just the same as plain addition, and that obviously weights the scale toward items with more points in their response format.

Even if the GHQ doesn't do that, the various Short-Form (SF) health measures that are widely used in the states certainly do, but then the people who devised them were not experts in measurement by any means.

Yes

Our analysis shows that there is not one underlying dimension and that 'negative' and 'positive' worded questions are answered quiet differently.

I have come from an (environmental) science background ( and also influenced by economics) where theory strongly guides measurement. It has been an eye- opener to see much common practice in the human sciences and reading the IRT literature and the debates over the Rasch model has been instructive.

read the following

• Johnson, Richard A.; Wichern, Dean W. (2007). Applied Multivariate Statistical Analysis (Sixth ed.). Prentice Hall. ISBN 978-0-13-187715-3.
• KV Mardia; JT Kent; JM Bibby (1979). Multivariate Analysis. Academic Press. ISBN 0-12-471252-5.

I think your 3 independent variables are categorized (3 locations) and continuous too. I will recommend is to convert the responses on the Likert Scale to continuous data. Then if assumptions permit you can go for multiple linear regression.

I would recommended to test your data in terms of the normality and homogeneity tests of variance, initially. Then you could to use the parametric or non parametric the statistical test. Due to the situation of datasets i think that it is better to use the one way ANOVA for parametric and the Kruskal-Wallis test for non parametric test.

I suggest you use MANOVA analysis

1. Multivariate Analysis of Variance

2.ANOVA

3. PCA

أفضل استخدام تحليل maniva لان عندك متعدد متغيرات

Multivariate analysis of variance (MANOVA) is an extension of the univariate analysis of variance (ANOVA).  In an ANOVA, we examine for statistical differences on one continuous dependent variable by an independent grouping variable.  The MANOVA extends this analysis by taking into account multiple continuous dependent variables, and bundles them together into a weighted linear combination or composite variable.  The MANOVA will compare whether or not the newly created combination differs by the different groups, or levels, of the independent variable.  In this way, the MANOVA essentially tests whether or not the independent grouping variable simultaneously explains a statistically significant amount of variance in the dependent variable.