Correlation with asymptotic standard errors

[berten]

How can one calculate the contemporaneous correlations for a group of variables with their significance indicated by their asymptotic standard errors in EViews? Thank you. It might really be a very simple issue. I tried the covariance analysis from group menu, and calculated the ordinary correlation coefficients with their t-values and p-values, but I see no option for asymptotic standard errors. Could someone clarify? Thank you.

[EViews Glenn]

Interestingly we'd never had a request for these. Most people are interested in the test statistics. EViews doesn't compute them automatically...I'll put it on the list to look at...

At first glance, with a little bit of work, you can compute the relevant terms using the SSCPs and other stuff that you can compute, but it's obviously not as easy as EViews doing the work for you. Note that you'll have to make a decision about whether to compute these under the null of no correlation or not.

[berten]

If the asymptotic standard error can be calculated from the statistics reported by EViews, would you mind giving an example? Thank you.

[EViews Glenn]

This isn't a particularly easy calculation in general and you'll need to make some assumptions about higher moments of r (see Kendall and Stuart, vol 4, p. 250). In the contingency table case (which is where I was looking earlier) there appear to be some simplifications, but the theory requires the contingency table setup.

Is there a reason that you need the asymptotic variances?