0
$\begingroup$

I'm doing a bit of business analysis at division level for a company. I'm interested in determining if there is a relationship between engagement (and other people metrics) and financial performance in 10-15 divisions of a company. This means I will have 10-15 rows of data. Is that enough data to do a correlation analysis?

I can double/triple the rows of data by including historical data, i.e., the same data for the 10-15 divisions for the previous Quarters of the financial year. Would this be incorrect to do?

Many thanks in advance for your help!

$\endgroup$

1 Answer 1

1
$\begingroup$

You can do a correlation analysis with N = 10 or 15, the problem is that the estimate will not be very accurate.

Adding historical data can bring in other problems, because the data are no longer independent. You can still run a correlation, but it's not so clear exactly what question you will be answering. If you decide to add the historical data, you might want to consider a more complex analysis that accounts for the dependence.

$\endgroup$
2
  • 1
    $\begingroup$ Thank you, this is very helpful. I was concerned about the dependence issue myself - do you have any recommendations on the other type of complex analysis I should consider? As for the small sample size, when you say the estimate will not be very accurate, does this mean the margin of error will be quite high? Many thanks again. $\endgroup$
    – Jade
    Feb 1, 2016 at 21:58
  • $\begingroup$ Multilevel models would be a good choice. And yes, a large margin of error is what I meant $\endgroup$
    – Peter Flom
    Feb 2, 2016 at 11:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.