VARMAResult.arch_test#

VARMAResult.arch_test(variable, lags=5)[source]#

Engle’s LM test for conditional heteroskedasticity in one equation.

Parameters:

  • variable (str) – Name of the equation to test.

  • lags (int) – Number of squared-residual lags in the auxiliary regression.

Return type:

_WaldTest

Returns:

The _WaldTest with lags degrees of freedom.

Raises:

SpecificationError – If variable is unknown or lags is outside 1..nobs-1.

Note

Rejection leaves the coefficients consistent but the reported covariance wrong, so Granger tests become unreliable while impulse responses stay interpretable as conditional means.