I have questions for my research. All data are ordinal.
1) What should come first? Correlation or multiple regression? (I am planning to use them both in my research).
2) Is it okay if my independent variables have a positive correlation with the dependent variable but after going multiple regression analysis some of my IV become insignificant ? (like having values less than the alpha value). How can I report this on my paper?
@pat.d could you please share the data here for an accurate answer. Thanks
It is customary for good reasons to report correlations first, typically in the form of a triangular table, listing the correlation of every variable against every other variable.
It is not at all unusual for a predictor variable to relate significantly with a criterion variable in correlation but lose the significant relationship in multiple regression. This could be a statistical artifact, e.g insufficient power because of too small a sample. It could also be meaningful in that the significant correlation reflects the fact that the particular predictor and the criterion correlate because both are correlated with one or more of the other variables in the study.
The comments by Paul Max Kohn are exactly correct. However, if your data are in fact ordinal, then it may be that neither (Pearson) correlation nor (OLS) multiple regression is appropriate. The questions you posted are so basic that I wonder if you should begin by seeking help (either at your university or through internet searches) on the fundamentals of statistical analysis and research design. ResearchGate might not be the best forum for obtaining the help you need.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression