PaigeMiller Tracker

Re: SAS Macros: Error Handling Best Practices

PaigeMiller — Thu, 10 Sep 2026 17:53:26 GMT

Turn on the MPRINT option to make reading the log (and figuring out what the error is) much easier.

options mprint;

Re: How to make the macro [works on column] to work on column with BY GROUP?!

PaigeMiller — Sat, 29 Aug 2026 18:34:22 GMT

@hellohere wrote:

P is the BYGRP

If @ballardw was here, he would say what I say to you now. Do not expect us to understand the problem at the same level as you do. If a variable is referred to in your text, and there is no variable that obviously has the same name, we are lost. If you want us to recognize that P is the variable you are talking about, then SAY IT IN YOUR PROBLEM DESCRIPTION. Don't make us ask.

Also, I assume you mean "P is the Group variable", but you haven't said that.

Re: How to make the macro [works on column] to work on column with BY GROUP?!

PaigeMiller — Sat, 29 Aug 2026 11:48:26 GMT

I have a macro. It works on column. The codes are below.

The real data need to work on is with BY GROUP.

The simulation is like below.

There is no GROUP variable here, so it is unclear what you mean. Can you please show us data that has a simulated GROUP variable?

It is also unclear why you think a macro is needed here. As far as I can tell, a macro is unnecessary and complicates the situation greatly.

The idea that you should perform calculations like this in SQL is a poor idea, that will lead you to trouble and inefficiencies, code will run slower and it will take you longer to program in SQL. (Just because you can do calculations like this in SQL doesn't mean you SHOULD do these calculations in SQL). Better to choose data steps or other PROCs to do calculations, especially with BY variables. The ability to handle BY variables is one of SAS's great strengths, you should endeavor to make use of BY variables in DATA steps and other PROCs, rather than via macros.

Re: sas base: working with results viewer window

PaigeMiller — Wed, 26 Aug 2026 14:15:22 GMT

If I am understanding you properly, and you have a long HTML output where you need to fix the 5th analysis, you need to re-run the analysis program that produced all of the long HTML output, making the fixes in there.

To explain what is happening, when you use ODS HTML (or other ODS destination), SAS creates a file containing all of the outputs you just created by sending everything to an HTML file (or other file). It also creates a set of indexes, which is what you are seeing on the left of the Results window, these let you navigate quickly to parts of the output HTML file. Deleting the index has no impact on the HTML file, it only prevents you from navigating there using the Results window. So if you want the fifth analysis in the HTML file to change, you have to fix the program that created the HTML file, and then re-run it.

Re: Graphics ODS Destination

PaigeMiller — Mon, 24 Aug 2026 19:06:36 GMT

&sasworklocation — is the for EG only? I don't seem to be able to use it in base SAS 9.4

Re: Graphics ODS Destination

PaigeMiller — Mon, 24 Aug 2026 18:36:51 GMT

I feel like I am missing something. The following code, using HTML, creates the plots as .png files in whatever folder you specify.

ods html path="C:\Users\xxxxxx\Documents\temp1" gpath="C:\Users\xxxxxx\Documents\temp1";
proc sgplot data=sashelp.class;
    scatter x=height y=weight;
run;
proc sgplot data=sashelp.class;
    vbar age/response=height;
run;
ods html close;

Re: SAS Support Communities metrics and their trends

PaigeMiller — Sun, 23 Aug 2026 10:56:28 GMT

It occurred to me today that not all forums have seen this decrease in traffic. The Adobe Community forums seem just as active and lively as ever. I have no statistics to back this up, but I use the Adobe forums almost daily.

Perhaps the difference is that SAS is a programming language while all Adobe products have a point and click interface. Or perhaps this is totally off the mark.

Re: Any way to rename a set of variable names?!

PaigeMiller — Tue, 18 Aug 2026 10:01:25 GMT

Quoting @Kurt_Bremser :

"why do you need to do this?"

Help us all out, which helps you as well ... explain to us the real problem you are trying to solve, not just the one piece of code where you are getting stuck. Normally, the real problem is that you are trying to create a table or a plot or analysis. People here have lots of experience to help you get there, but we need to know where it is you want to go. See the XY problem.

Re: I want to vertically expand the plot

PaigeMiller — Tue, 11 Aug 2026 10:45:43 GMT

Please explain in a lot more detail what you mean by "I want to vertically expand the plot".

Re: Strange behavior when data are written and re-read directly from an Excel file

PaigeMiller — Fri, 07 Aug 2026 11:37:22 GMT

Length and format are two different characteristics of your variable named X. Just because you have set the length of X to be $200, the format of X has not been set, so SAS guesses at what you want, and in this case it guesses wrong. If you set the format in PROC REPORT, the problem goes away.

proc report data=xxx; 
columns x;
define x/format=$200.;
run;

or, an alternative solution, with PROC REPORT unchanged from your original code

data xxx;
  length x $200;
  format x $200.;
  x = '-1'; output;
  x = '-10'; output;
  x = '-100'; output;
  x = '-10'; output;
  x = '-1'; output;
run;

PS: maybe I'm going off on a tangent here, but storing numerical values as text strings is generally a problematic thing to do, while storing numerical values as numbers is highly recommended, and also solves the problem.


data xxx;
  x = -1; output;
  x = -10; output;
  x = -100; output;
  x = -10; output;
  x = -1; output;
run;

ods excel file="&workpath.\XXX.xlsx" style=Plateau;
ods excel options (sheet_name="XXX" tab_color="white" flow="tables");

/* Even simpler would be to use PROC PRINT here */
proc report data=xxx; 
    columns x;
    define x/display;
run;

ods excel close;

libname prvfndgs xlsx "&workpath.\XXX.xlsx";

data yyy; 
  set prvfndgs.XXX;
run;

Re: Transpose Account Data which has multiple entries into multiple columns.

PaigeMiller — Tue, 04 Aug 2026 11:31:50 GMT

You can use PROC TRANSPOSE as in this example:

https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/proc/n01rzqgzs8vq1bn10h1wtx914kyf.htm

If you want actual code, you will need to provide data in a usable form, as seen in these examples and instructions.

PS: I understand that a key person wants the data in this form; but please understand that is the only reason to do this. If you are programming for yourself, and want to use the results in further analyses, tables or plots, you would never do this transpose. As stated by @Kurt_Bremser , this transposed arrangement of the data makes all other programming much more difficult.

Re: SAS Support Communities metrics and their trends

PaigeMiller — Mon, 03 Aug 2026 13:55:13 GMT

@FreelanceReinh wrote:

Yes, it looks like those mathematical inaccuracies occurred too frequently in Copilot's training data.

Sadly, it's not just using words wrong. It doesn't do calculations properly either. One day I was lazy and asked Copilot to compute MPH for my walk, in which I walked 3 miles (exactly) in 56 minutes and 36 seconds. Copilot walked me through the math and for the first step, Copilot told me that 56 minutes and 36 seconds is 56.61 minutes. Can you spot the error? (it's a small error but certainly not round-off error) There was another math error in the next calculation as well.

Re: SAS Support Communities metrics and their trends

PaigeMiller — Sun, 02 Aug 2026 22:06:21 GMT

I asked an AI chatbot (Copilot) why traffic is way down at internet help sites. The answer:

Traffic on StackOverflow and similar help sites is down because several structural shifts all hit at once, and the data is now overwhelming: generative AI replaced most routine Q&A, moderation friction pushed newcomers away, and expert contributors withdrew when their effort stopped being rewarded. The collapse is not subtle — question volume is down 29× from its peak, and many sites are back to launch‑year activity levels.

Interesting how Copilot does not seem to recognize that multiplication and division are not the same thing. Down 29x is not the same as 1/29th. I do realize that many people make this same mistake in their speech and in their writing.

Re: Will this work to partition a large data set into 10?

PaigeMiller — Fri, 31 Jul 2026 16:30:05 GMT

It would still help if you would be specific, instead of using the very vague word "processing". I'm sure many people here have good ideas on how to handle large data sets, but we can't help you unless we know what you going to do next.

Also, how big is this data set, in terms of rows and columns?

Re: Will this work to partition a large data set into 10?

PaigeMiller — Fri, 31 Jul 2026 15:53:35 GMT

I would like to suggest that splitting data sets like this is rarely needed and is rarely a good idea, and it usually makes the next steps in your programming more difficult. If you tell what you plan to do next with this data, we could advise on a better approach.

If, for example, you want to perform a regression analysis on each split, you can do so like this (and of course this method applies to almost every PROC even if it is not regression)

data large_set1;
    set large_set;
    id=mod(_n_,10);
run;
proc sort data=large_set1;
    by id;
run;
proc freq data=large_set1;
    by id;
    model y=x;
run;

Re: What is collinearity and why does it matter?

PaigeMiller — Mon, 27 Jul 2026 17:58:26 GMT

Partial Least Squares ought to be mentioned here. In my opinion, it is the best modeling tool when you have collinearity among the X variables, and is widely used in some areas, like spectroscopy (and of course is used in many other areas). Please see the paper by Tobias (of SAS Institute), in which he takes 1,000 highly correlated predictors, and still obtains a useful model, without all the time-consuming messiness of reducing variables. Also see the paper by Bartell (from JMP) here.

Re: html email

PaigeMiller — Mon, 20 Jul 2026 10:52:25 GMT

Perhaps the %NRSTR function will work here. But yes, definitely show us your code. In fact, please, from now on, ALWAYS show us your code in the first post on a topic.

Re: Sum values over repeated variable values

PaigeMiller — Fri, 10 Jul 2026 17:54:02 GMT

Agreeing with @Tom that numbers should not characters.

Once you perform the conversion to have numeric values, PROC SUMMARY or PROC MEANS produces the dataset or output you want.

Re: Estimation and significance test

PaigeMiller — Wed, 08 Jul 2026 13:42:23 GMT

SAS has never (as far as I know) used the normal distribution when there are more than 30 data points. It uses the t-distribution even when there are 30,000 data points (or more) assuming of course that the t-distribution is the correct distribution for the problem.

The idea of using the normal distribution when there are more than 30 data points is an approximation, and was useful back in the days before computers and you had to use a look-up table. But it isn't really needed with modern computers.

Re: Estimation and significance test

PaigeMiller — Tue, 07 Jul 2026 18:56:53 GMT

Are you using PROC VARMAX? It would appear that the parameter estimates follow a t-distribution.

According to the documentation, the parameter estimates converge to a multivariate normal distribution as T approaches infinity (T apparently is the number of data points), which means before they converge they are likely a t-distribution.