MCLEODLI—procedure for McLeod-Li test for nonlinearity

[TomDoan]

@MCLEODLI is a procedure for doing the McLeod-Li test for nonlinearity (ARCH effects). It's a commonly used diagnostic on the standardized residuals from a GARCH model to test for remaining ARCH effects. This test is actually quite simple (it's just a CORRELATE applied to the squares of the series), but we provide this for convenience. McLeod and Li(1983), "Diagnostic Checking of ARMA Time Series Models Using Squared Residual Autocorrelations", J. of Time Series Analysis, vol 4, pp 269-273. Note that you apply the procedure to the standardized residuals, not to their squares. (The procedure handles the squaring).

Detailed description

[Vik_pa]

Dear Tom,

I have read that McLeod Li test is used on the residuals from the conditional mean model. And to check the ARCH effect, Li-Mak test is used on the squared standardized residuals obtained from the GARCH model.
Li-Mak (1994) "On the squared residual autocorrelations in non-linear time series with conditional heteroscedasticity" described it.
So is it possible to use the Li-Mak test in RATS.

Thanks

[TomDoan]

There's a reason that obscure 20 year old papers are obscure. The Li-Mak test is not a relatively simple single test—it's complicated and different for every type of GARCH model. The McLeod-Li test has stood the test of time as a quick and easy check for a gross failure of an estimated GARCH model. If a more exacting test gave a .04 significance level rather than a .10, what would you do differently? Nothing.