def test_gwas_linear_regression__scalar_vars(ds: xr.Dataset) -> None:
res_scalar = gwas_linear_regression(
ds, dosage="dosage", covariates="covar_0", traits="trait_0"
)
res_list = gwas_linear_regression(
ds, dosage="dosage", covariates=["covar_0"], traits=["trait_0"]
)
xr.testing.assert_equal(res_scalar, res_list) # type: ignore[no-untyped-call]
def run(traits: Sequence[str]) -> Dataset:
return gwas_linear_regression(
ds,
dosage="dosage",
covariates=["covar_0"],
traits=traits,
add_intercept=True,
)
def test_gwas_linear_regression__lazy_results(ds):
res = gwas_linear_regression(ds,
dosage="dosage",
covariates="covar_0",
traits="trait_0",
merge=False)
for v in res:
assert isinstance(res[v].data, da.Array)
def test_gwas_linear_regression__variable_shapes(ds, chunks):
ds = ds.chunk(chunks=chunks)
res = gwas_linear_regression(ds,
dosage="dosage",
covariates="covar_0",
traits="trait_0",
merge=False)
shape = (ds.dims["variants"], 1)
for v in res:
assert res[v].data.shape == shape
assert res[v].data.compute().shape == shape
def _get_statistics(ds: Dataset, add_intercept: bool,
**kwargs: Any) -> Tuple[DataFrame, DataFrame]:
df_pred: List[Dict[str, Any]] = []
df_true: List[Dict[str, Any]] = []
for i in range(ds.dims["variants"]):
dsr = gwas_linear_regression(
ds,
dosage="dosage",
traits=[f"trait_{i}"],
add_intercept=add_intercept,
**kwargs,
)
res = _sm_statistics(ds, i, add_intercept)
df_pred.append(dsr.to_dataframe().rename(
columns=lambda c: c.replace("variant_", "")).iloc[i].to_dict())
# First result in satsmodels RegressionResultsWrapper for
# [t|p]values will correspond to variant (not covariate/intercept)
df_true.append(dict(t_value=res.tvalues[0], p_value=res.pvalues[0]))
return pd.DataFrame(df_pred), pd.DataFrame(df_true)