EViews.com

I have an (unbalanced) panel dataset. The cross section is organised by state (with many observations for each state) and the date series is year of marriage. I need to construct a heterogeneity index which captures (average) heterogeneity in various spousal differences by state by year. I can do this using the view by classification option but since I have a limited number of observations for some years it returns NA or 0 standard deviations. My solution is to calculate the standard deviation for observations over a 5 year period (i.e. the year of marriage, the two years prior and the two years after). I have tried using the @mstdev and @movstdev commands - for example @mstdev(education,5) but this is not providing correct results as it is not calculating the standard deviation by state or year. Is there any way that I can use the statistics by classification option but get it to compute the standard deviations on a 5 year rolling basis rather than for single years?

Thanks in advance for any advice.

The only way to do it is to program it yourself by rolling the sample forward.

Within EViews if you click on Help->Quick Help Reference->Sample Programs and Data, and then click on Roll, there are some examples of computing a rolling standard deviation.