Sorry I been very busy probably like most of us, but I want to thanks everybody for support. I saw a few times the 'maxims' document rolling around, I will definitely read it, and if it is not one of them, I would add to sleep a day on the problem if you get stuck, fresh ideas will appear! It may takes a few days but I will get back with more details asap ... View more

Very interesting, I can't wait next day to start trying things! It feels like the break time will be busy about trying! Thanks Everybody for all good ideas, next step is to understand what happens and fix it properly, not to just make it works with a patch! ... View more

I will check what happen with option xsync, I use usually set options to noxwait, I suspect that it could be something as simple as that... What ever I find as interesting result, I will share! ... View more

I use a 'call system' for encrypting with 7zip, I don't know about encrypting with ods package but I will check at it, thanks a lot for the idea! ... View more

Thanks for that inspiring answer, I will investigate few things in my code/log related to your message. I'm just surprised that encrypting works when automated in 2 separated macros, that makes me think that 7zip path is working and it doesn't get stuck in junk... I'll keep you in touch with result, thanks again! ... View more

Hello Koen, just to share results :   I tried with proc report but I wasn't able to calculate statistics by rows e.g. row proportion... I don't mind about the look of the report, I need numbers to be all within a single page ready to use. Here is what I finally obtained with proc sql and proc tabulate, it can handle any 2 dimensions table size with any kind of stats included :   data status; input row_group $ col_group $ value proportion tout $; datalines; 0$-15$ 0%-40% 10 30 all 26$-35$ 41%-60% 34 41 all 16$-25$ 41%-60% 22 51 all 26$-35$ 41%-60% 26 54 all 16$-25$ 61%-100% 17 90 all 26$-35$ 0%-40% 32 10 all 26$-35$ 61%-100% 30 78 all 26$-35$ 41%-60% 26 45 all 16$-25$ 0%-40% 22 12 all ; run; /* total global */ proc sql; create table status2 as select *, count(tout) as n_obs, mean(value) as tot_val_mean, mean(proportion) as tot_prop_mean from status; quit; proc sort data=status2 out=tot nodupkey; by tout; run; data tot; set tot; row_group="total"; col_group="total"; rename tot_val_mean=val_mean; rename tot_prop_mean=prop_mean; rename n_obs=freq; pct=100; col_pct=100; row_pct=100; drop value proportion tout; run; /* total colonnes */ proc sql; create table status3 as select *, count(tout) as col_freq, 100 as col_pct, count(tout)/n_obs as pct, 100 as row_pct, mean(value) as col_val_mean, mean(proportion) as col_prop_mean from status2 group by col_group; quit; proc sort data=status3 out=col_tot nodupkey; by col_group; run; data col_tot; set col_tot; row_group="total"; rename col_freq=freq; rename col_val_mean=val_mean; rename col_prop_mean=prop_mean; drop value proportion tout N_obs tot_val_mean tot_prop_mean; run; /* total lignes */ proc sql; create table status4 as select *, count(tout) as row_freq, 100 as row_pct, count(tout)/n_obs as pct, 100 as col_pct, mean(value) as row_val_mean, mean(proportion) as row_prop_mean from status3 group by row_group; quit; proc sort data=status4 out=row_tot nodupkey; by row_group; run; data row_tot; set row_tot; col_group="total"; rename row_freq=freq; rename row_val_mean=val_mean; rename row_prop_mean=prop_mean; drop value proportion tout N_obs tot_val_mean tot_prop_mean col_freq col_val_mean col_prop_mean; run; /* total cellules */ proc sql; create table status5 as select col_group, row_group, count(tout) as cell_freq, count(tout)/N_obs as pct, count(tout)/col_freq as col_pct, count(tout)/row_freq as row_pct, mean(value) as cell_val_mean, mean(proportion) as cell_prop_mean from status4 group by col_group, row_group; quit; proc sort data=status5 out=cell_tot nodupkey; by col_group row_group; run; data cell_tot; set cell_tot; rename cell_freq=freq; rename cell_val_mean=val_mean; rename cell_prop_mean=prop_mean; run; data tout; set cell_tot col_tot row_tot tot; run; proc tabulate data=tout; class col_group row_group; var freq pct col_pct row_pct val_mean prop_mean; table row_group*(freq pct col_pct row_pct val_mean prop_mean), col_group; run;       code isn't cleaned but it works, i felt like recoding proc freq the way i would like it ahah!   Thanks again and sorry if my english isn't as good as if it would be natural to me!   Marc       ... View more

Thanks again Koen, it comforts me that intuition was good! I'm mostly using proc report to make lists, but I did know it was much more powerful... With that documentation, I have everything to customize my output as needed. I was probably trying too much to create a matrix style result and simply display it but indeed I can put "raw" data in proc report and tell it what to do with them!   Have a very good day!   Marc ... View more

Thanks Koen for the fast answer! Problem isn't to get the stats, it is to display them... Multiple tables are by groups, my output need to be .rtf or .xlsx file ready for user in this format but 10x10 classes :    Rows_group Lines_group Frequency Percent Row percent Col percent Value mean Proportion mean I was thinking to build 'manually' the table and display results with proc report or proc tabulate but it was getting tricky, I ended thinking it wasn't the good approach. Proc freq gives me most important stats user need, but I been asked if I can add means of variables which created the classes. and I like to please users who have good use of stats!!!   I'm still trying to find an idea but I'm stuck in circular reasoning and worse case it will end in 2 separate tables but it wouldn't be nice and finished job...   Thanks again!   Marc   ... View more

Thank you Tom! I will be able to fit that small change in the program, and we shouldn't have any similar problems in the future! Thanks again!!   ... View more

@Ksharp wrote: Could you write some data step code to dynamically pad the blanks before format's LABEL for the dataset you are used to generated format ? after that using "PROC FORM CNTLIN=dataset " to make the right format ? That was the plan, scanning the 'mother' table and changing 'É' to 'E' if in first position, it would be acceptable for the program results. But lets say categories are persons names, Émmy and Emmy would be problematic, and it would need to be shown in good order among the others.    I'm not sure using blank spots would fix it, it would always bring all the E É in good order but in top of all categories. Removing them after creating the format doesn't fix the problem when using order=formatted in proc tabulate. ... View more

Thank you ballardw for the time spent helping me! You are right, to be able to give appropriated solution (and optimize testing scenarios) you need to know which are the constraints of the actual format I use. It will take me some time to answer with precision, the format creation is a bit in a black box in the program, but I won't give up! ... View more

In this simple exemple it works fine if there are no categories starting with A B C D. But it needs to be more robust...   In reality, the format is dynamic, created from another SAS table being udpate automatically. In that table, order is made by variable value (AAA, BBB, ... as exemple), labels are not necessarily in order ('AAA'='Taxe' BBB='Étudiant'). If I just patch that table, it won't prevent for any further problems caused by new categories.   We see that for accent characters not in first position, alphabetical sort is working fine by default. Problem is when first character is an accent character, and it might be a tiny limit of using SAS with french, it would be very rare that 2 categories would be exactly same letters but first with accent character or not. It could happens in names lists in which accents are very important but I never noticed same problem with proc report. ... View more

Changing that in the format step of the code i added in first post show the problem more clearly : proc format; value $ctype "AAA" = "E(*ESC*){unicode '0301'x}n" "BBB" = "En" "CCC" = "Invé" other = "Inve"; run; The rule says e (E) comes before é (É), it display fine for Invé that comes after Inve. But if we force with unicode accent, order is not really correct because Én comes before En, and it is not good order according to writing rules...   Sorry for the editing, I was lost myself 😄 but problem persists! ... View more