Logistic Regression Sample Size

Select 'Logistic regression' in G*Power for your analysis.

Decide between one-tailed or two-tailed based on your hypothesis.

Choose α = 0.05 for a 5% risk level of type I error.

Aim for a test power between 0.80 and 0.99 for reliability.

For simple models, R² of other variables is zero.

Choose the distribution for X: Binomial, Normal, or Poisson.

Use pilot study data for accurate parameter estimation.

Input X's mean and SD from pilot data for precision.

Use PSPP to find Exp(B) for variable X from pilot data.

Calculate the likelihood of y=1 under the null hypothesis.

Combine parameters in G*Power to find the needed sample size.