I'm looking to produce some phi correlations and I'd rather not do it manually, but I can't find a library for R that includes phi.
Given Greek letters are used by lots of folks in lots of different ways, you might say which phi you are referring to. Be useful to give the formula.
If you mean the Cramer's V for a 2x2 table, or just the correlation, say.
Hi Alex,
There is a function named "phi" in psych package. Here the link.
https://www.rdocumentation.org/packages/psych/versions/2.0.12/topics/phi
I hope it helps you..
Depends. For the correlation coefficient you can compute from the chi square statistic and n:
e.g.,
phi.from.chi
Daniel Wright Hüseyin Yıldız Babak Jamshidi Thom Baguley
Hi all,
Thank you for your responses.
I'm using a 2x2 contingency table (image attached) with the dichotomous variables of National Identity Assigned and Cause of Problem (0 = No, 1 = Yes in both variables). I was using Phi (which was returned at -.01) given the dichotomous variables.
In R:
> tab phi.from.chi
> chisq.test(tab, correct=F)
Pearson's Chi-squared test
data: tab
X-squared = 0.049462, df = 1, p-value = 0.824
> tab chisq.test(tab, correct=F)
Pearson's Chi-squared test
data: tab
X-squared = 0.049462, df = 1, p-value = 0.824
> phi.from.chi phi.from.chi(0.049462, sum(tab))
[1] 0.01123292
So Phi is 0.01123292 and the sign just depends on how you code the categories (i.e., what is 1 and what is zero)
Thom Baguley That's what I got when I ran phi manually (attached image) and later chi-square through R. I also installed the suggested libraries and got the same results. So everything seems to be aligning.
- Badges
- Science topic
- Similar topics
- Psychology