0

Why in regression analysis does the inclusion of a new variable make statistically insignificant coefficients that previously were not? In model 3 length of contraception becomes insignificant when controlling for age. How to interpret this

Description of three models

EngrStudent
  • 8,232
  • 2
  • 29
  • 82
Devansh Gupta
  • 11
  • It is likely that age is a confounder of your variable of interest. That's often the reason for change in significance level when extra regressors are included. – Caserio Feb 27 '17 at 20:14

1 Answers1

0

This might mean that there is high correlation between length of contraception and age. Now it could just by chance also, so if you have access to more data I would take that.

Check the VIFs. If VIF is large (alteast greater than 5, then multicollenearity exists and produces large SE and low p - value, hence producing wide confidence interval bands). If you have multicollnearity you can:

  1. Leave it as is. The predictions will not be biased (will only be imprecise)
  2. Remove Length of contraception and check how much SSR is reduced. If not by much this new equation could be a solution.
  3. Trying standardizing variables
anuragsodhi
  • 56
  • 6