6
$\begingroup$

I am currently self-studying statistics and I'm confused about the null model in binary logistic regression. I understand that the null model is used to be compared with the model you designed, but what exactly is the null model? Just ln(x)=y?

$\endgroup$
2
  • 5
    $\begingroup$ It is an intercept only model, where the only parameter is related to the proportion of '1' in the population. $\endgroup$
    – Michael M
    Commented Jan 21, 2014 at 18:47
  • $\begingroup$ Probably not just $\ln(x)=y$, unless you've defined those variables rather unusually. (And please do define variables in questions.) $\endgroup$ Commented Jan 21, 2014 at 19:19

1 Answer 1

13
$\begingroup$

The full model is $$\ln \frac {\pi}{1-\pi}=\beta_0 +\beta_1 x_1 +\beta_2 x_2+\ldots$$ where $x_i$ is the $i$th predictor, $\beta_i$ its coefficient, & $$\pi=\Pr(Y=1)$$ where $Y$ is the response (coded 1 for "success" & 0 for "failure")

The null model, as @Michael says, contains just the intercept: $$\ln \frac {\pi}{1-\pi}=\beta_0$$ So the intercept is the log-odds of "success", estimated without reference to any predictors.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.