Solved – Testing for heteroskedasticity of time series in R

Question

I wish to test my time series data for volatility clustering, i.e. conditional heteroskedasticity.

So far, I have used the ACF test on the squared and absolute returns of my data, as well as the Ljung-Box test on the squared data (i.e. McLeod.Li.test).

In a recent paper (http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1771862, the test is reported on page 8) co-authored by a well-known researcher, they have employed the White test (http://ideas.repec.org/a/ecm/emetrp/v48y1980i4p817-38.html) to directly test for heteroskedasticity.

I have tried the same approach, however was unable to do so.
From my understanding, the White test needs residual variance (usually from a linear regression model) as an input.

Now my question is: How did the researchers perform the White test? I do not understand which inputs they used for their White test.

While searching for solutions, I have found the sandwich package which uses the vcovHC and vcovHAC functions to estimate a heteroskedasticity-consistent covariance matrix, however the input is also a fitted linear regression model..

Answer 1

You should try Mcleod Li Test for heteroskedasticity available in R's TSA package. That package can take just time series as input.