Enum RegressionError 

pub enum RegressionError {
    EmptyInput {
        what: &'static str,
    },
    ShapeMismatch {
        what: &'static str,
        expected: usize,
        got: usize,
    },
    NoResidualDegreesOfFreedom {
        n: usize,
        p: usize,
        df: isize,
    },
    RankDeficient,
    InvalidParameter {
        msg: String,
    },
    InvalidResponse {
        msg: String,
    },
    NotConverged {
        iterations: usize,
        msg: String,
    },
}

Errors returned when constructing an OlsFit or computing a diagnostic on it.

The policy on degenerate inputs is deliberate and documented per-variant: the crate hard-errors on inputs where no meaningful statistic exists (empty data, a shape mismatch, non-positive residual degrees of freedom, a rank-deficient design matrix) rather than silently returning NaN. The one place NaN is used intentionally is a per-column result slot that is genuinely undefined for that column (e.g. the VIF of the intercept column); that is a documented sentinel, not an error condition.

Variants

EmptyInput

The design matrix or target was empty.

Fields

what: &'static str

Which input was empty.

ShapeMismatch

Two inputs that had to agree on a dimension did not (e.g. X has a different number of rows than y has entries).

Fields

what: &'static str

Human-readable description of the mismatched quantity.

expected: usize

The value that was expected.

got: usize

The value that was supplied.

NoResidualDegreesOfFreedom

Residual degrees of freedom (n - p) are not strictly positive, so the residual variance, and every statistic derived from it, is undefined.

This is exactly the “n close to the number of parameters” edge case: with n <= p the model has no residual freedom left and reporting any residual-based diagnostic would be misleading, so construction fails clearly instead.

Fields

n: usize

Number of observations supplied.

p: usize

Number of model parameters (design-matrix columns).

df: isize

The offending n - p value.

RankDeficient

The design matrix is not full column rank, so the OLS solution is not unique and QR cannot recover the coefficients.

Perfectly collinear predictors are the usual cause. VIF has its own, softer handling of near/exact collinearity (it reports a very large or infinite value rather than erroring); this variant is for the primary model fit, where a non-unique solution has no sensible fallback.

InvalidParameter

A caller-supplied hyperparameter was outside its valid range (e.g. a negative ridge/lasso penalty λ).

Fields

msg: String

Human-readable explanation of what was wrong.

InvalidResponse

The response passed to a logistic fit was not a valid binary outcome — either it contained values other than 0 and 1, or it was entirely one class (so the maximum-likelihood fit is degenerate / non-identifiable).

Fields

msg: String

Human-readable explanation of what was wrong.

NotConverged

An iterative fit (IRLS for logistic regression, coordinate descent for lasso) did not converge within its iteration budget.

For logistic regression the usual cause is perfect or quasi-complete separation, where the maximum-likelihood coefficients diverge to ±∞ and no finite fit exists — a real modeling problem the caller needs to know about, not a solver detail to paper over.

Fields

iterations: usize

Number of iterations attempted before giving up.

msg: String

Human-readable note on the likely cause.

Trait Implementations

impl Clone for RegressionError

fn clone(&self) -> RegressionError

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for RegressionError

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Display for RegressionError

fn fmt(&self, __formatter: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Error for RegressionError

fn source(&self) -> Option<&(dyn Error + 'static)>

Returns the lower-level source of this error, if any.

fn description(&self) -> &str

👎Deprecated since 1.42.0:

use the Display impl or to_string()

fn cause(&self) -> Option<&dyn Error>

👎Deprecated since 1.33.0:

replaced by Error::source, which can support downcasting

fn provide<'a>(&'a self, request: &mut Request<'a>)

🔬This is a nightly-only experimental API. (error_generic_member_access)

Provides type-based access to context intended for error reports.

impl PartialEq for RegressionError

fn eq(&self, other: &RegressionError) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for RegressionError