Hi,  If I have this dataset (psedo, it is really boig instead) data data; input DATE date9. A $ B; format Date date9.; datalines; 19NOV2008 100000011 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 24NOV2008 100000022 10 ; run;  You see that they have the same status (10=deafult let's say) and ChangedDate is differ.. If I want to remove all the "duplicates"  they're not duplicates beacuse ChangeDate differs for every row. But I want to have the minimnum date of ChangeDate beacuse it sounds more logial, since that company was already in the status deafult from the begging of my dataset but how can I only keep on row of the company = '100000022 ' without nodup/nodupkey/distinct ?  Thanks!   ... View more

Hi,  I'm so stuck and wondering if someone could please help me 🙂   I have this data that looks like this;  For year = 1990 rate score fauilere %     A 0,0001 0,00011 B 0,0002 0,00022 C 0,0003 0,00033 D 0,0004 0,00044   For year = 1990+n rate score fauilere %     A 0,0001 0,00011 B 0,0002 0,00022 C 0,0003 0,00033 D 0,0004 0,00044     For year = 1990+(n+1) rate score fauilere %     A 0,0001 0,00011 B 0,0002 0,00022 C 0,0003 0,00033 D 0,0004 0,00044   How can I best plot this to see the changes to see allt of this, I'm so slow? In Excel preferbly (otherwise SAS Base)     ... View more

Hi I try to calculate a MWB; In mathematics, particularly matrix theory, a band matrix or banded matrix is a sparse matrix whose non-zero entries are confined to a diagonal band, comprising the main diagonal and zero or more diagonals on either side.   And it requires a nxn-matrix ,but my data looks like this:  And I think the best way to calculate it when the matrix don't have the same dimensions is to add a nullvector, in my case when Mk1=scale1    data pseudo_data; input SCALE $ mk2 mk3 mk4 mk5 mk6 mk7; datalines; scale1 0.620073 0.585710 0.742480 0.654074 0.152773 0.828327 scale2 0.030940 0.081121 0.505554 0.495800 0.024913 0.874568 scale3 0.493812 0.464746 0.675302 0.900584 0.240976 0.112287 scale4 0.194194 0.483504 0.650082 0.795761 0.112740 0.145614 scale5 0.427196 0.986343 0.778883 0.631170 0.623107 0.586194 scale6 0.037071 0.375161 0.217043 0.893637 0.779019 0.161867 scale7 0.186883 0.830461 0.038186 0.000778 0.336085 0.586100 ; run; /* ex rclassorgcount dataset */ data rclassorgcount; input SOMETHING_THAT; datalines; 100 200 150 250 300 350 400 ; run; %let rcmigrdist=mk2 mk3 mk4 mk5 mk6 mk7; /* # of mk's*/ %let K=7; /*scale*/ %let K_Max=10; /*how many mk's it could possibly be depending on the dataset you running*/ %let SOMETHING_THAT=SOMETHING_THAT; /*something else in my real code; count(SOMETHING) as SOMETHING_THAT*/ proc iml; use pseudo_data; read all var {&rcmigrdist.} into rc_migr_dist; /* Load matrix with movements shares p_ij */ close pseudo_data; use rclassorgcount; read all var {&SOMETHING_THAT.} into rc_orig_count_vec; /* Load vector with number of humans per movement i, Ni */ close rclassorgcount; /* Calc Matrix_normal_u */ M_norm_u = 0; do i = 1 to &K. - 1; sum_p = 0; maxval = max(abs(i - &K.), abs(i + (&K_Max. - &K.) - 1)); N_i = rc_orig_count_vec[i]; do j = i + 1 to &K.; sum_p = sum_p + (rc_migr_dist[i, j] / 100); /* Assuming percentages are given, convert to decimal */ end; /*print sum_p;*/ temp = maxval * N_i * sum_p; M_norm_u = M_norm_u + temp; end; create M_norm_u var {"Matrix_normal_u"}; append from M_norm_u; close M_norm_u; quit;       And you see that the log says:    153 end; ERROR: (execution) Invalid subscript or subscript out of range.   How can I add the nullvector for mk1 in this case?   there is more matrix operations  then later  M_norm_u Upper MWB UMWB_skalad M_norm_l Lower MWB LMWB_skalad       0 0 0 A B C         (A,B,C is calculated numbers that the matrix could calculated but not the other where there is 0)   (i dont post it here) beacuse I think there should be a step before all the actual calculations, like before the <actual> proc iml-step? Thanks! ... View more