Hello Margarita,

Much of the answer depends on your specific research aim(s) and question(s). If no single independent variable shows a significant correlation with the DV, then a regression model isn't going to be significant either (unless you have an extraordinary example of a suppression effect going on amongst your variables). Thus, you won't find out anything different from the regression.

However, if your research question concerned the efficacy of a regression model for explaining differences in your DV, then run it and report the (nonsignificant) results.

I will add that a sample of 37 cases is on the low side for a regression analysis in general, and one with as many as 4 IVs in particular. You don't have a lot of statistical power to detect relationships unless they are strong. An N of 37 yields a power of about .38 (worse than a coin flip) of detecting a "medium" effect size (using Jacob Cohen's guidelines; an R-squared of about .13 ) with 4 IVs.

Note: These values presume that Pearson correlations are used. Spearman correlations are just Pearson correlations computed from ranked values.

Good luck with your work.

The bivariate correlations examine bivariate relations. The individual ordinal regression coefficients, if you use all four variables, are about conditional relationships. Which of these are your research hypotheses about?

To add to the conversation, it's ok if you find a null result. Stopping data collection at targeted positions (data peaking) is sometimes called p-hacking. A null result tells us something important as well as a significant result. Also, did you pre-register your study?

Daniel Wright Examine bivariate relations. More info: Variable 1 (x) consist of the combined value of variables A, B, & C. My goal is to test this value's relationships to the dependent variable (y). Subsequent tests will then be done with each variable A, B, & C against y.

David Morse Thank you. Just for clarification, there are 37 companies over a 5 year period. So there are 185 observations. I think it would prudent to consider increasing my sample size.