ProbitLink<T>
Models & Types · AiDotNet.LinkFunctions
Probit link function: g(μ) = Φ⁻¹(μ), where Φ is the standard normal CDF.
Implements: ILinkFunction<T>
For Beginners
Probit is similar to logit but uses the normal distribution instead of the logistic distribution. Key differences:
- Probit: Based on normal distribution (bell curve tails)
- Logit: Based on logistic distribution (slightly heavier tails)
In practice, results are usually very similar. Probit is sometimes preferred when:
- The underlying process is modeled as a latent normal variable
- You want consistency with other normal-distribution-based methods
Interpretation: The coefficient represents the change in z-score for a one-unit change in the predictor.
How It Works
The probit link maps probabilities to z-scores from the standard normal distribution. It’s an alternative to logit for binary classification.
Constructors
| Constructor | Summary |
|---|---|
ProbitLink() |
Properties
| Property | Summary |
|---|---|
bool IsCanonical { get; } |
|
string Name { get; } |
Methods
| Method | Summary |
|---|---|
T InverseLink(T eta) |
|
T InverseLinkDerivative(T eta) |
|
T Link(T mu) |
|
T LinkDerivative(T mu) |
|
T Variance(T mu) |