This question already has an answer here:

In a (logistic?) regression paper related to credit scoring and involving psychometric data and credit scoring, there are 11 predictor variables used to estimate credit score. The sample size is 142. Intuitively, it doesn't seem enough, but what exactly is the rule of thumb involved for regressions? That involve psychometric data? That are logistic?

the paper: An empirical investigation of dispositional antecedents and performance-related outcomes of credit scores.

ZUI slideshow summary: An Empirical Investigation of Dispositional Antecedents and


marked as duplicate by Scortchi - Reinstate Monica Nov 20 '15 at 10:49

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • 4
    $\begingroup$ If may seem small before you do the calculations. But after calculations it may turn out 142 sample is big enough. The power also depend on the effect size. $\endgroup$ – Deep North Sep 18 '15 at 13:49
  • $\begingroup$ There is a bunch of various phenomena called Curse of Dimensionality. The best explanation I've seen is here. $\endgroup$ – Alexander Lutsenko Sep 18 '15 at 14:22
  • 1
    $\begingroup$ My English is so bad, whenever I write comments I will find grammar mistakes later. $\endgroup$ – Deep North Sep 18 '15 at 14:41
  • $\begingroup$ As the minority class could number 71 at most, you should certainly be alert to the possibility of over-fitting - was the model validated? If you look at Harrell (2001), Regression Modelling Strategies, p61 there are references to some empirical studies justifying the 10-20 events per variable rule of thumb. $\endgroup$ – Scortchi - Reinstate Monica Nov 20 '15 at 10:52