The solution that I have come up with, that approximates the solution you want, is to create a data table that has concatenated to it, a copy of the complete data table for each of the years. For the first of the concatenated copies, you set the year value to the first year in your data table, then for the second, you set it equal to the second year, etc. For each of the concatenated copies, you set the rowstate to Hide. The end result of this is that for the calculation of the density, all of the data will be used, but the individual data points will not be displayed. Only the rows from the original table, which the rowstate is not hidden or excluded will be displayed.
The above solution can be made more exact, if the first of the concatenated data tables contains all of the rows from the original data table, except for the first year, the second concatenated table has all of the rows except for the second year, etc. It is a bit more complex to produce but marginally more accurate.
Jim