Calculate proportion of expression variation explained by genotypes (cis-heritability; \(h_g^2\)).

Parameters:

X: Array | ndarray | bool_ | number | bool | int | float | complex¶

\(n \times p\) matrix for independent variables with no intercept vector.

y: Array | ndarray | bool_ | number | bool | int | float | complex¶

\(n \times 1\) vector for gene expression.

covar: Array | ndarray | bool_ | number | bool | int | float | complex | None = None¶

\(n \times m\) matrix for covariates.

normalize: bool = True¶

Boolean value to indicate whether normalize X and y

Returns:

A tuple of

1. genetic variance (float) of the complex trait,

2. \(h_g^2\) (Array) from limix definition,

3. LRT test statistics (float) for \(h_g^2\),

4. LRT \(p\) value (float) for \(h_g^2\).

Return type:

Tuple[float, Array, float, float]

Example

Estimate cis-heritability for a gene:

import numpy as np
from sushie.utils import estimate_her

# Genotype matrix (100 samples, 500 SNPs)
X = np.random.randn(100, 500)
# Gene expression
y = np.random.randn(100)

# Estimate heritability
g, h2g, lrt_stat, p_value = estimate_her(X, y)
print(f"Heritability: {h2g:.3f}, p-value: {p_value:.4f}")