I currently work on my bachelor's thesis and analyse the correlation between 13 countries over a time period of 117 years. I built a correlation table and now want to say somethin about the average correlation over time. I divided the full sample into three subperiods of different size. Now two questions came up:

  1. Is it possible to simply compute the mean for each period by averaging the single correlation coefficients? Or is it analogue to this: Working with correlation coefficients

  2. When Building the weighted average of the three subperiods, should it equal the average of the full sample? (currently not the case)

  • $\begingroup$ You estimated the correlation between each pair of countries for the 117 years, this is your "mean" correlation. Now you want to do the same thing, except that you will divide the data into periods? $\endgroup$ Aug 8 '19 at 9:26
  • $\begingroup$ My Intention is to build the average correlation across countries. I am wondering if this is done by simply computing the mean of the pairwise correlations derived before. $\endgroup$
    – Anton20
    Aug 8 '19 at 9:33
  • 1
    $\begingroup$ Some people do not like that, stats.stackexchange.com/questions/8019/… $\endgroup$ Aug 8 '19 at 9:35
  • $\begingroup$ Thank you that helped! However, the so computed averages for the subperiods do not equal the average of the full sample, is this correct? $\endgroup$
    – Anton20
    Aug 8 '19 at 11:06
  • $\begingroup$ Could you explain how you might interpret such an "average correlation over time"? That might help us understand what kind of an average you are looking for. $\endgroup$
    – whuber
    Aug 8 '19 at 12:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.