0

I'm writing a report on housing vacancy, examining various combinations of variables for factors correlated with vacancy rates recorded in the census.

This is for a simple "requires further investigation" section of the report and all I want to do is present some scatter plots with a descriptive of the correlations.

I'm not interested in super-population arguments; I'm not looking to perform inferential stats here; I just want an appropriate metric for the reader.

Looking at other postings here it seems this is always a point of contention, including using Pearson's. Could someone give me some theoretical basis for why or why not listing Pearson's as a descriptive statistic here would be appropriate? And if not, what other correlation might be appropriate and why?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
leixlipred
  • 53
  • 3
  • 2
    This thread has a reasonable discussion of 'correlation' including a focus on the limitations of Pearson correlation metric. https://stats.stackexchange.com/questions/179511/why-zero-correlation-does-not-necessarily-imply-independence/179551#179551 – Mike Hunter Nov 27 '17 at 13:06
  • Thanks DJohnson, hadn't seen that thread in my searches. Will read now – leixlipred Nov 27 '17 at 13:17

0 Answers0