Sure! So, the first table below is what I have: People were tested twice, and for each testing time there is a separate row for each person. Test scores are different between rows within a person but demographic variables are the same.    ID Time Gender Race Age Test1 Test2 Test3 1 1 0 1 22 50 70 50 1 2 0 1 22 60 75 45 2 1 1 3 31 35 60 30 2 2 1 3 31 80 65 60 3 1 1 5 18 25 85 50 3 2 1 5 18 45 90 40 4 1 1 1 42 10 20 80 4 2 1 1 42 60 70 50 5 1 0 2 21 50 60 60 5 2 0 2 21 55 40 80 6 1 1 2 36 70 40 15 6 2 1 2 36 80 45 30   The second table is what I want: to have one row per person and to have two variables for each test -- one variable for the test at Time 1 and another variable for the test at Time 2. And like that for each test.   ID Gender Race Age Test1_1 Test1_2 Test2_1 Test2_2 Test3_1 Test3_2 1 0 1 22 50 60 70 75 50 45 2 1 3 31 35 80 60 65 30 60 3 1 5 18 25 45 85 90 50 40 4 1 1 42 10 60 20 70 80 50 5 0 2 21 50 55 60 40 60 80 6 1 2 36 70 80 40 45 15 30 ... View more

So, this code works only if I have 1 test. But now I have many tests... So, I thought maybe I can do this code for all of my test variables separately and then just merge the data sets.    Here is what I came up with so far. The problem is that I don't know how to assign names to the "want" data set so that each newly created want has its unique name, e.g., want_Test1, want_Test2, etc. (So that I can merge them at the end by the ID and demographic variables). The same problem with COL1 and COL2 that are in each want data set -- I need to rename COL1 into Test1_1 and COL2 into Test1_2 in the first data set, COL1 into Test2_1 and COL2 into Test2_2 in the second data set, etc. Could someone please help me with that?   data varset_test; length test_name $40; do test_name = 'Test1', 'Test2', 'Test3'; output; end; run; %macro trans(test=); proc transpose data=have out=want(drop=_:); by ID Race Gender Age; var &test; run; %mend trans; data new; set varset_test; call execute('%trans(test='||strip(test_name)||');'); run; Thank you in advance.    P.S. Maybe there is an easier way to do transposing for multiple test variables than what I am trying to do?  ... View more

It works! Thank you so much! ... View more

It works! Thank you!!   A follow-up question: I am not quite sure which of my variables are attributes of a person (i.e., don't vary within a person between Time 1 and Time 2) and which are attributes of a test (different values for Time 1 and Time 2 for the same person). Is there an easy way to find out which variables are which?    The best option I found so far is to split the file into two: one file with Time 1 values and another file with Time 2 values, and then use proc compare. It gives me a table with unequal values... but I thought maybe there is a way that I can get a LIST of variables with unequal values and a LIST of variables with equal values so that it's easier to copy-paste in proc transpose? (I have about 600 variables, so I am really looking to automatize this process...).   Any ideas would be highly appreciated! ... View more

THANK YOU! I knew there got to be an easy way to do that. 🙂 ... View more

Thank you for your reply. My data look something like this:   ID   Cluster Var1 Var2 1          1        2       3 2          1        1       5 3          2        4       4 4          2        6       2 5          2        1       3 6          3        4       1 7          4        4       6 8          4        7       3 9          4        5       5 10        4        2       1 ...   Each case (as indicated by ID) is clustered. For example, Cases 1 and 2 are clustered within Cluster 1; Cases 3, 4, and 5 are clustered within Cluster 2, Case 6 is the only case in Cluster 3; and Cases 7-10 are clustered within Cluster 4. I also have values for each case on Var1 and Var2.   What I want to do is to select randomly 1 case per cluster. So, after selection, I will have a data set of 4 cases, each belonging to a unique cluster. After that, I will use this newly formed data set to work with Var1 and Var2, e.g., test for mean differences between Var1 and Var2.    I understand how to do repetitive work -- I can just have a loop. Also, the RAND statement, I believe, will not help me as I am not generating data -- I already have data from which I just need to randomly select 1 case per cluster.   Thank you in advance. ... View more

That makes sense. Thank you, Rick! ... View more

Thank you! I think I figured out how it works. I didn't use a grid for now, and I also didn't use the RandFleishman function (it creates a matrix, and I haven't figured out yet how to use it). But I think the code below works correctly (and I am sure it looks very clumsy but I am just learning...). But I am curious about why my skewness and kurtosis values have such a big range and variance (especially kurtosis)... I thought in all samples these values will be quite close to the specified.    /* Define and store the Fleishman modules */ %include "C:\...\RandFleishman.sas"; %let NC = 500; %let N_samples = 20; %let skew = 1.5; %let kurt = 3; proc iml; load module=_all_; /* load Fleishman modules */ c = FitFleishmanFromSK(&skew, &kurt); /* find Fleishman coefs */ create DD var {"c"}; append; close; quit; proc transpose data=DD out=FF; run; proc sql; create table GG as select COL1 as c1, COL2 as c2, COL3 as c3, COL4 as c4 from FF; quit; data MC (keep = sampleID X); set GG; call streaminit(12345678); do sampleID = 1 to &N_samples; do i = 1 to &NC; N1 = rand("Normal"); X = c1 + c2*N1 + c3*(N1**2) + c4*(N1**3); output; end; end; run; proc univariate data=MC noprint; by sampleID; var X; output out = OUTSTAT mean = mean std = SD skewness = skew kurtosis = kurt min = min max = max; run; proc print data = OUTSTAT; run;   ... View more

Thank you! Did you mean Section 16.8 Fleishman's Method? I am using an online version of the book through my university's library, and it doesn't have page numbers.    I am slowly figuring things out... So, I understood that I need to save the RandFleishman.sas file, create a new sas file, and call for the RandFleishman.sas file there. I read the code that you provide in Section 16.8 in the book but, as far as I understood, it fits the model to the data. But I don't have data. I just need to generate a distribution with the specified skewness and kurtosis. How should I change the code in this case?    Thank you again.  ... View more