Skip to main content
. 2014 Mar 20;9(3):e90257. doi: 10.1371/journal.pone.0090257

Table 4. Summary of the steps in constructing the multinomial regression model using the backward elimination method.

Model Action Effect(s) −2 Log Likelihood χ2 for removala df p
0 Entered All effects 794.42
1 Removed Proportion of Younger Brothers 794.42 0b 0b
2 Removed Total Number of Siblings 795.26 .84 2 .66
3 Removed Proportion of Younger Sisters 798.76 3.50 2 .17
a

The χ2 for removal is based on the likelihood ratio test.

b

Because each proband’s proportions of older brothers, older sisters, younger brothers, and younger sisters is necessarily summed to 1.00, these proportions were perfectly multicollinear. To reduce the multicollinearity, the computational algorithm of the SPSS multinomial logistic regression program eliminated the proportion of younger brothers from the set of predictor variables.