2

I have a dataset, for example, of success factors (columns) noted by several subjects (rows). Basically, when a subject notes a factor, its marked as a yes or no.

I want to do an analysis to find factors that seem to be related or appear together. correlation analysis. I'm thinking of giving the yes/no values either 1 or 0 respectively and performing covariance on that with excel. Is this a valid approach?

Will appreciate any comments on this

Mohamed Omar
  • 21
  • 1
  • Computing the Pearson correlation $r_{ij}$ on each $(i,j)$ pair of columns would help you identify associated factors (in the sense of tending to be 0 together or 1 together). In fact $nr_{ij}^2$ is exactly the chi-square for the corresponding $2\times 2$ table for that pair of $0$-$1$ variables. – Glen_b Sep 13 '13 at 08:51
  • Specifically, the Pearson correlation on a pair of $0$-$1$ variables is the same as the [phi coefficient](http://en.wikipedia.org/wiki/Phi_coefficient). – Glen_b Sep 13 '13 at 09:03

0 Answers0