liveBook · Manning

McFadden's pseudo R-squared

As explained in Statistics Slam Dunk

McFadden's pseudo-R2 is a goodness-of-fit metric for generalized linear models based on the log-likelihoods of a fitted model and a null model.

McFadden’s pseudo-R² evaluates logistic-regression goodness of fit by comparing a fitted model’s log-likelihood with a null model’s. Higher values indicate greater predictive power; values around 0.2 or 0.4 can indicate a good fit. section 14.6.2

NBA playoff qualification

For a logistic model predicting playoff qualification from points allowed, run pR2(fit3)["McFadden"]. The result, 0.3084425, supports treating the predictor as a good fit. A competing points-scored model returns 0.06446806, making the first predictor stronger by pseudo-R². section 14.6.2

How Statistics Slam Dunk covers it

How deep it goes
Defines the log-likelihood comparison, gives interpretation thresholds, and demonstrates calculation without deriving the formula.
Examples
Uses NBA playoff qualification models based on points allowed and points scored, returning pseudo-R² values of 0.31 and 0.06.
What it assumes
Assumes logistic regression and generalized linear models, while distinguishing pseudo-R² from ordinary linear-regression R².
What you can do after
Apply it via pscl::pR2() to compare logistic-regression fit and identify the stronger predictor model.
How it connects
Connects logistic regression with generalized linear models, null models, log-likelihood, AIC, and predictive-model comparison.
Cover of Statistics Slam Dunk
liveBookmanning.com
Read Statistics Slam Dunk on liveBook Get the book at manning.com

Also explained in

Compare how 2 books explain McFadden's pseudo R-squared →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Statistics Slam Dunk. The section links point to the parts of the book it draws on.