I was hoping someone could help me on a query about correlation on spss. Thanks so much. :)?

Question

I have a sample of 15 firms, with certain data i.e. the profits for each firm, the number of employees, the number of board members etc. Due to the small sample size, its not worth doing multiple regression to see whether the the number of employees, the number of board members etc. affect the firms profits. So I just want to do some correlation between the dependent variable (profitability) and the independent variables i.e. member of employees, the number of board members etc......should I use Pearsons or Spearman? Thanks.

Answer 1

For metric data you should use Pearsons.

Spearman is only good if you have rankings.

Answer 2

Take the average profitability and dichotomize your dependent as above/below average profitability or do a median split. Once your outcome is dichotmous, you can do an exact logistic regression. This is acceptable given your sample size. Add your predictors one at a time and look for the AIC optimal model. Or run regular old MLE logistic looking for the AIC Optimal. This can be done in one step by using an entry criteria of .95 and a stay criteria of .99. Output the parameter estimates and fit statistics. Looking at the fit statistics when AIC reaches a nadir the corresponding number of variables is what you want in the model. You can then shop around this number by running models with 2 or 4 predictors. During the shopping phase use Exact Logistic Regression with Monte Carlo simulation.

Answer 3

Ivo is right. Spearman's is for rank correlation only, and it looks like you are going to correlate the data. You do have a small sample size, though, so I might consider removing one degree of freedom in the Pearson product moment formula, using n-1 rather than n. Just a thought.