Hi,   to my knowledge, in order import data and work on it in SAS the data has to be in table form structure, i.e, varaible headers and observations in columns.   Recently I was working with data that had a very "un-table" like structure, and importng it into SAS was a mess. In the end I was able to transform its data into table form using Excel VBA, by working with cells and finding their structure pattern. But Excel/VBA is limited and sometimes I witnessed it go crazy when asked to think too much.   Nevertheless the code to transform the data structure was fairly easy because of the grid-nature of Excel, and a grid is somewhat like a matrix, so I was wandering if proc iml has data structure transforming capabilities like VBA, if not better.     Thank you ... View more

Hi,   I am doing many regressions on the the data set sashelp.cars.   I have created a table with all the models for which I want to do the regressions, and this table has a combination and count variables that look like this:   Combination count Horsepower MPG_City Weight Length 4 EngineSize Length 2 Cylinders Horsepower Length 3 … …   then I created macro variables &comb and &num_iv for each of the combinations   Then I did the following macro which should do all the regressions for all combinations that have >= 3 variables:   %macro all_regressions; ods exclude all; ods output parameterestimates = parameter_estimates; ods output anova = anova; ods output FitStatistics = fit_statistics; proc reg data=sashelp.cars ; %do k = 1 %to &cnt; %if &&num_iv&k >=3 %then %do; m&k: model invoice msrp = &&comb&k; %end; %end; run; %mend; %all_regressions; quit; At first everything seemd to work well but then I notices something:   In my ANOVA table, the Total DF for all the models = 425. This means that 426 observations were used. But the data sashelp.cars has 428 observarions. I did a proc means and it showed me that the var "Cylinders" is the only one that has 426 observations instead of 428, and upon observation I saw that these are obs= 248 and obs=249 which it is missing.   But the problem is that because the models which include the var Cylinders have 426 observations, SAS assigns 426 observatins to ALL the other models, even though they have 428 observations. In this case 2 observations won't make a big difference, bt it does make a difference when it is hundreds and thousands of observations.   Is there a way to solve this problem?     Thank you!  ... View more