CODEWITHSUNDEEP
rmachine-learningstatisticslinear-regression
Harish
Harish

Reputation: 119

Impact of a predictor becomes opposite in multiple logistic regression model

When I am modeling my dependent variable(continuous) on single independent variable(continuous), the predictor is impacting the dependent variable in a positive way. But, when I am modeling the same dependent variable on multiple independent variables, the independent variable which impacted positively when modeled individually is impacting negatively in the final model with several independent variables.What could be the reason for this?Any views would be of great help.

Upvotes: 0

Views: 141

Answers (1)

slonopotam
slonopotam

Reputation: 1710

This might happen when your 'independent' variables are in fact correlated with each other. Simplified example:

y = x2-x1
x2 = 2*x1+e
which means y = x1+e (where e is random noise)

Learning regression model y=f(x1) will give us positive weight for x1, while learning y=f(x1,x2) will likely give negative weight for the same variable.

Upvotes: 2

Related Questions