You may choose some related classifier for your data at hand. Have a look at Article Supervised classification with interdependent variables to s...

https://rstudio-pubs-static.s3.amazonaws.com/347013_371e59e8867549ddb922ea7b530e5c6a.html

Analysis of variance (ANOVA) will compare the four means. If the ANOVA is statistically significant, a post-hoc technique like Duncan's multiple range test might be used to separate the variables.

Muhammad Ali thank you very much for your answer, i will read the paper and see if it will be helpful

Ette Etuk Thank you for your answer. But the issue is that i do not have a dependent variable which is needed for ANOVA. I am working with four simple independent variables you may say which are measured through a single question each. And there were four surveys given out which means i want to compare the four datasets and the same variable with each other in the different datasets. I hope I am clear on this

The first consideration is if you will combine your five question into a single scale. And if so, if it's reasonable to treat this scale as a (continuous-like) interval variable. ... As opposed to 5 separate ordinal measurements.

Salvatore S. Mangiafico Thank you for your answer, and the questions will not be combined as it is a manipulation check and tests for a different aspects of the scenario provided before answering the questions. So every question stands on its own simply to see the differences between the four treatment groups (4 different surveys with different scenarios).

In that case, I would say the ideal approach is probably ordinal regression. With modern software, this won't be any more difficult than conducting a common anova, although it may be less familiar, and maybe more difficult to find good guidance depending on the software. ...

Otherwise, a Kruskal-Wallis test, followed by an appropriate post-hoc test (maybe Dunn 1964 test) will actually work fine. Provided you are using an implementation that handles ties correctly. (Most software implementations are fine with this, but ties may not handled right if you are trying to do it by hand, or maybe with cheesy software.)

Do you have 5 different questionnaires or 5 different questions? If you have 5 different questionnaires you can try cluster analysis, to try and group your questions based on similarity.

If you only have 5 different questions, and 4 response variables, it's different.

You mention you have 5 independent variables, are those your responses? If they are, we call them dependent because they need to be associated with your independent variables, the predictors.

However, to rigor even the predictors can't be called independent as at times they may present correlation (e.g. colinearity).

Hello Muhammad Taqi, I completely agree with my professor; Prof. Ette Etuk. You can use the ANOVA to compare 4 groups, however, your data is ordinal (data from a Likert scale) you need to transform the data from all five questionnaire to a discrete data. If there exist a significant difference, then you may need further analysis (a post hoc test) to know which group was significantly different from which.

Godwin Lebari Tuaneh Jose Risomar Sousa Salvatore S. Mangiafico

Thank you guys for your response, ok so I will clarify everything here because it seems that I have not been clear enough. I have also attached one of the surveys which are related to the economic treatment group.

Research Design: Quasi-Experimental Design

Survey: Scenario-based

Number of Questionnaires: 4 (1 for each group)

Treatment groups: 4 Treatment groups

• Economic animosity
• Political animosity
• War animosity
• People animosity

So the questions in the surveys are all same except for the one related to the specific animosity dimension (economic, political, war, and people).

So I have attained the data, now I want to compare these surveys to see how each of the questions has a different impact due to the animosity dimension and the scenario given.

I am checking based on the questions:

How can I compare the data from four surveys? There is no grouping variable that I can use for the t-test. And there is no dependent variable that I can use for ANOVA. I have checked the descriptives but I need to put forth statistical significance in the thesis.

Thank you

Hi Dr Muhammad Taqi . See the following link: https://www.researchgate.net/post/What-is-the-best-statistical-test-to-compare-4-variables

See also the following link: https://stats.stackexchange.com/questions/112163/how-to-compare-4-independent-means-when-there-is-no-dependent-variable

Also see the following RG link: Article Statistical Methods for Comparison of Data Sets of Construct...

Aref Wazwaz Thank you very much for the material and the links. First link you mentioned, there everyone has mentioned to conduct ANOVA but in order to conduct ANOVA there is a need of the factor which needs to be added for the comparison. What would I add for that part is the problem.

You are most welcome Muhammad Taqi . Wish you the best.

Muhammad Taqi , I don't really understand what you are trying to determine with your analysis. ... Maybe look at the correlations between the Animosity dimension and each of the other questions, for each survey group ? You might want to use Kendall correlation for this (maybe Tau-b because of the ties). You could get confidence intervals for the Tau statistic to see if there are differences among the surveys. ... I don't know if this would address what you are trying to find out.

Hello Muhammad Taqi,

You said there is no grouping but there is, your treatment groups are the grouping. code it as

Economic animosity = 1

Political animosity =2

War animosity =3

People animosity=4

convert the responses on each questionnaire item by all the respondent to a single discrete value through its average value or sum. so that you will have

1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4 vertically as your grouping variable (factor) for the ANOVA, then; the discrete values from each questionnaire say;

4.5, 6.8, 8.1, 5.5 for Economic animosity 5.2, 3.9, 4.6 5.1 for Political animosity and so on. also arrange that vertically and you have values for your animosity dimension impact. This is okay for a one way ANOVA. I hope this is clear enough

As follow up to the above answer, the number of respondent is also important for an accurate degree of freedom

Godwin Lebari Tuaneh Thank you very much for your response and I think now it becomes much more clear. I have sent you a message, could you please have a look at it. Thanks

https://www.researchgate.net/post/What-is-the-best-statistical-test-to-compare-4-variables

Muhammad Taqi, the best way is to fix 3 parameters and change one, record the data, and change the second, till you finish all the parameters. Another method is using the Taguchi method to select some values in each parameter.

Check https://libguides.library.kent.edu/SPSS/CompareMeans

Also check http://jjcweb.jjay.cuny.edu/akaci/files/chapter_91_comparing_means-t-tests_and_anova.pdf