Hi all, I would like to fill in missing values using compound interest rates and I would like SAS to continue compounding for a certain number of periods beyond the scope of the current data set. My problem is that I have variables that are missing at different points in the data set. For example, Have: date     gdp     gdp_rate     empl     empl_rate 1          10          1.004          20          0.998 2            .           1.004          30          0.998 3            .           1.004          40          0.998 4            .           1.004            .           0.998 Want: date     gdp     gdp_rate     empl     empl_rate 1          10          1.004          20          0.998 2        10.04       1.004          30          0.998 3        10.08       1.004          40          0.998 4       10.12       1.004        39.92       0.998 5        10.16       1.004        39.84       0.998 6        10.20       1.004        39.76       0.998 Here is a sample of the do loop I am using. It works fine if I only have one observation in the dataset (to get this I fill in the missing values and keep only the last observation), but I would like the do loop to fill in the missing observations and then keep going until the last iteration. The problem is that the do loop is repeating for each observation... data want;      set have;      *starting values for temporary vars;      _gdp=gdp;      _emp=empl;      *compute moving estimates by compounding interest;      do i=1 to 27;           _gdp + gdp_rate*_gdp - _gdp;        *retains new estimate; *new estimate is actually just the rate times the current value, but this is the only way I knew to retain it;           _empl + empl_rate*_empl - _empl;           new_gdp=lag(_gdp);                       *lags new estimate so it can be used to replace missing values and fill in future values of gdp;           new_empl=lag(_empl);           if i>1 then                do;                     gdp=new_gdp;                       *sets gdp to new estimate computed using previous period gdp times the growth rate;                     empl=new_empl;                end;           output;                                             *outputs observation;      end; run; Again, this works fine if there's only one observation in the input data set (e.g. if I just kept observation where date=1). If there are more observations in the input data set as above, the do loop iterates for each observation (i.e. 27 times for date 1, 27 times for date 2, etc.). Any ideas how to fix this? Or, perhaps you know a much easier way to accomplish this? ... View more

I am trying to compute quarterly moving averages within a by group; however, there are different numbers of observations within each group. I have seen some examples, but they assume the same number of observations need to be averaged throughout. Here is an example: Items a and d are purchased in each quarter, but item b is only purchased in the second and third quarters, while item c is only purchased in quarter 4. So, the by group would be "by item quarter" item     quarter     quantity a          1               2 a          2               1 a          3               4 a          4               3 b          2               25 b          3               4 c          4               15 d          1               2 d          2               1 d          3               4 d          4               3 The example above is sort of a watered down version. I am using time series data and even though an item is purchased every quarter in one year, it may not be purchased every quarter in another year, or it may not be purchased at all for a particular year. Is there a simple way to compute moving averages for this type of data without interpolation? If possible, can you also include an example with interpolation? Any help would be appreciated. Thanks! ... View more