I have 7 variables (1 interval, 4 ordinal and 2 nominal) with levels between 4-9, for which responses are received through a survey. My question is how I can find the relation between these through statistical analysis? I want to use 3 to 6 variables as dependent and 1 as an independent variable.
You might be interested in association rules in the WEKA toolkit Article Usage Apriori and clustering algorithms in WEKA tools to min...
or https://forums.pentaho.com/threads/60228-Finding-Similarities-in-Data-using-Weka/
Dear Shamsher,
When making inferences, it would be useful to know what will be your main outcome measure. Having 3-6 variables as dependent with one as independent seems a little strange. It sounds like you have 6 outcome measures. It would be more appropriate to be the other way round. But it certainly depends on your study objectives and main outcome measure which we need to look at first. You do have the options to analyze those variables using conventional statistical methods, or the most robust type ones through machine learning algorithms. The point now is, what is your main outcome measure? Variables in principle can simply be classified as continuous and categorical. Determining your outcome measure will direct to the nature of the variables, yet points to the appropriate statistical test that we need to do.
Shamsher Singh Multiple correlation, multiple logistic regression, and factor analysis can be used.
Kurubaran Ganasegeran Sir, you are right. I have made a mistake in question drafting. I have 1 dependent variable and 3-6 independent variables. I am looking for statistical tests for the same. Can you name some tests?
It now depends on the nature of your dependent variable, and subsequently your independent variables. If your dependent variable is continuous in nature, and if you wish to find association with a continuous variable as well, you may perform correlation analysis (Pearson correlation). If your independents have two categories, you may use independent t-test, three categories of more, use ANOVA with post-hoc test. In contrast, if your dependent is categorical (two categories) and your independents have two categories as well (use Chi-square), if three or more please direct your self to binary logistic regressions. If you have more than two categories for your dependent, you may enter into more complex regressions, such as ordinal regression. These are some common examples that may direct you to some idea. There are plenty more statistical tests that could be considered, all depending on the nature and type of your variables.
First, univariate analyses must be carried out for the corresponding variables (CHi² or T-test). Then the correlation must be calculated, for example with Spearmns rho. The relation between the variables can be shown with the logistic regression analysis. One possibility with many variables would be to add the independent variables to the dependent variable in blocks in order to evaluate them statistically.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis