Weighted frequency plots

[rdwest]

Hi,

I'm sure this is a very elementary question for anyone who is experienced in graphical analysis using Eviews: I need to make distribution plots of data where the relative frequency is indicated by a series rather than by the number of records in the data set. So, for example, I'd like to make density plots or histograms of x (which is not really a continuous variable, but takes on many sequential discrete values) against its frequency, which would be indicated by a weighting variable y times the frequency of observations.

Sorry if this is elementary, but I didn't see a "weighting" option or the like for any of the distributional plots, and I failed to find an appropriate description in the help files.

Any ideas?

Many thanks,
R.

[EViews Glenn]

Unfortunately, EViews doesn't support frequency weighted data in very many places (really, only in GLM estimation at the moment).

[rdwest]

Thanks, Glenn.

I'm not looking for estimation procedures, though -- but rather graphical diagnostics. I need to create a categorical version of the variable of interest: grouping a variable that ranges from 1 - ~500 into several logical groups according to the underlying frequency distribution within the data. (Eventually, this will feed into a multinomial discrete choice model). Since I'll have to define these categories myself, I want to make sure my delineations reflect meaningful breaks in the data. Since the data set I'm working with is very large, the best way I know to make such determinations is to look at the relative frequency of each value. However, there's a lot of noise when the dependent variable values are compared on one-step intervals, so it would be really useful to have a kernel density plot or a histogram, or other ideas for visualizations that would smooth over this noise a bit.

Thanks!
Rachel