# What particular measure to use? Multiple regression or MANOVA?

I've been doing a research paper - Effects of Workplace Bullying on Employees' Productivity, Self-Confidence and Self-Esteem. So I have 1 IV (Workplace bullying) and 3 DVs. what can I use to interpret it?

Another thing: if it's MANOVA, can I execute it without having to use factors?

• What does this mean? "So I have 1 - IV (Workplace bullying) and 3 DVs" 3 discrete variables? What is "1-IV (Workplace bullying)". Making your question clearer will help to boost the chance of your getting a good answer. – Simon Hayward Sep 4 '13 at 9:16
• Presumably 1 IV means one independent variable and 3 DVs means three dependent variables. Abbreviations and acronyms should be explained, please: see meta.stats.stackexchange.com/questions/1479/… – Nick Cox Sep 4 '13 at 9:57
• Yes at one point it read '1 - IV' and i wasn't sure if there were 4 categories of bullying or 1 independent variable... – Simon Hayward Sep 4 '13 at 10:20

You have three outcomes and one input variable, you can't use multiple regression. Peter has clearly explained, you need to choose between three simple regression (taking one output at a time) or MANOVA (Multivariate regression).

Regression techniques to be used, if the output is continuous:

1. Single input & single output - Simple regression (If input categorical, use dummy variable or go for t-test/ANOVA)

2. Multiple inputs & single output - Multiple regression (If input(s) categorical, use dummy variable(s) or go for ANCOVA)

3. Single input & multiple outputs - Multivariate regression

4. Multiple inputs & multiple outputs - Multiple multivariate regression

And if the outcome is categorical, then it becomes classification problem. Different methods can be used like logistic regression, discriminant analysis etc.

• Thanks @amoeba. I've edited. I should have rechecked before posting. – Dr Nisha Arora Mar 16 '16 at 9:42
• I question the correctness of your definition of MANOVA and MANCOVA in points 3 and 4, if what you wrote implied definition. – ttnphns Mar 16 '16 at 10:44

Whether you want a) MANOVA b) Multivariate regression or c) several OLS regressions d) something else depends on what you want to test and the nature of the data. If you are interested in relationships among the three dependent variables (DV) as well as between the DVs and the single independent variable (IV) you want either MANOVA or multivariate regression (the latter not to be confused with multiple regression, which has one DV and more than one IV).

ANOVA and regression are really the same model, but the ANOVA/MANOVA terminology is usually used when your independent variable is categorical and the regression/multivariate regression when the IV is numeric/continuous.

You also have to consider the nature of the DV: All the above assume it is continuous. If it is not, then you probably want some form of logistic regression.

• what i really want to prove is that Bullying has nothing to do or does not have an effect on the 3 dependent variables that i used. what particular regression can prove this? we are using statistica software so it is really difficult on my part to execute such computation since i am really not that good in statistic. what result should i focus on aor look at and use for interpretation.thanks – lindsey Sep 4 '13 at 13:54
• Please answer the questions - they don't require any computation. 1) Are you interested in the relationship among the DV or not? and 2) Which variables are categorical and which are not. – Peter Flom Sep 4 '13 at 14:51
• 1) no i'm not interested in the relationship of the dv. 2) all of the variables are not categorical – lindsey Sep 4 '13 at 15:48
• Then you can use 3 regression analyses. – Peter Flom Sep 4 '13 at 15:53
• what particular regression analysis is needed for this kind of paper? – lindsey Sep 4 '13 at 16:02

IV is independent variable, DV stands for dependent variable (output). So in this case you indeed have one independent (input/cause) variable: the amount of bullying and 3 outcome variables which are studies (i.e. Productivity, Self-Confidence and Self-Esteem). A regression analysis should be applied to test for effects.

MANOVA would be used to test multiple groups and although your question does not clearly state this, I am going to assume you have one group of persons on which measured the amount of IV and tested all 3 DVs, so MANOVA is not the thing you need.

In this case you could also choose to test for correlations between variables instead of using a regression technique, i.e. to correlate the IV with the 3 DVs, although now you use these variables as random variables.

• In the first paragraph you show why some form of regression should be used; in your latter two you say it shouldn't be. Correlation does not distinguish outcome variables from independent variables. – Peter Flom Sep 4 '13 at 11:08
• Yes you are right. But, I think the question is rather unclear so I mainly wanted to point out what the abbreviations in the text were meaning as to clarify. Furthermore, I do still think the question is about comparing different variables, such that correlation should be used. Maybe you could clarify yourself @Linsey ? – Fraukje Sep 4 '13 at 11:31
• Both regression and correlation (and other methods) compare variables. If you have IVs and DVs (and the question is pretty clear about that) then some form of regression/ANOVA analysis is probably what is needed. – Peter Flom Sep 4 '13 at 11:32
• Oh i see the question has been edited, it is more clear now what is asked and I agree with you that a regression technique should be used. – Fraukje Sep 4 '13 at 12:01
• hi, you have given me a clear explanation about what i really need to do. thank you very much for the help. yes iv means independent variable and dv means dependent variables. the research were conducted within the group wherein each participant have taken "all" the scales. i'll use the multiple regression. thanks thanks thanks. :) – lindsey Sep 4 '13 at 13:28