liveBook · Manning

Linear predictor

As explained in Timeless Algorithms, The Foundational Ideas

A linear predictor is a weighted additive combination of input variables and coefficients that produces an intermediate model score.

In logistic regression, the linear predictor combines input variables and coefficients into a score: z = β₀ + β₁x₁ + ⋯ + βₖxₖ. It is not the predicted response; it represents the log-odds of the positive outcome. The sigmoid function then maps this score to a probability between 0 and 1. section 3.3.2

z=β0+β1x1+β2x2+⋯+βkxk

The additive form makes each predictor’s weighted contribution to the score explicit. section 3.3.2

A score of zero

With an intercept of −6, coefficient 1.2, and input x = 5, the score is zero; the corresponding predicted probability is 0.5. section 3.3.2

How Timeless Algorithms, The Foundational Ideas covers it

How deep it goes
Defines the weighted sum, interprets it as log-odds, and connects it to the sigmoid transformation.
Examples
Evaluates a score using an intercept of −6, coefficient 1.2, and input value 5.
What it assumes
Reviews probability and odds, defining odds before interpreting their logarithm as the predictor.
What you can do after
Understand how a weighted score represents log-odds and maps to a predicted probability.
How it connects
Links logistic regression to log-odds, odds, and the sigmoid function.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Timeless Algorithms, The Foundational Ideas. The section links point to the parts of the book it draws on.