phi: Find the phi coefficient of correlation between two dichotomous variables

Description

Given a 1 x 4 vector or a 2 x 2 matrix of frequencies, find the phi coefficient of correlation. Typical use is in the case of predicting a dichotomous criterion from a dichotomous predictor.

Usage

phi(t, digits = 2)

Arguments

a 1 x 4 vector or a 2 x 2 matrix

digits

round the result to digits

Value

phi coefficient of correlation

Details

In many prediction situations, a dichotomous predictor (accept/reject) is validated against a dichotomous criterion (success/failure). Although a polychoric correlation estimates the underlying Pearson correlation as if the predictor and criteria were continuous and bivariate normal variables, and the tetrachoric correlation if both x and y are assumed to dichotomized normal distributions, the phi coefficient is the Pearson applied to a matrix of 0's and 1s.

The phi coefficient was first reported by Yule (1912), but should not be confused with the Yule Q coefficient.

For a very useful discussion of various measures of association given a 2 x 2 table, and why one should probably prefer the Yule Q coefficient, see Warren (2008).

Given a two x two table of counts llll{ a b a+b (R1) c d c+d (R2) a+c(C1) b+d (C2) a+b+c+d (N) } convert all counts to fractions of the total and then \ Phi = [a- (a+b)*(a+c)]/sqrt((a+b)(c+d)(a+c)(b+d) ) =\ (a - R1 * C1)/sqrt(R1 * R2 * C1 * C2)

This is in contrast to the Yule coefficient, Q, where \ Q = (ad - bc)/(ad+bc) which is the same as \ [a- (a+b)*(a+c)]/(ad+bc)

References

Warrens, Matthijs (2008), On Association Coefficients for 2x2 Tables and Properties That Do Not Depend on the Marginal Distributions. Psychometrika, 73, 777-789.

Yule, G.U. (1912). On the methods of measuring the association between two attributes. Journal of the Royal Statistical Society, 75, 579-652.

Examples

Run this code

phi(c(30,20,20,30))
phi(c(40,10,10,40))
x <- matrix(c(40,5,20,20),ncol=2)
phi(x)

Run the code above in your browser using DataCamp Workspace