About Rick_SAS

Rick_SAS · ‎03-07-2012

Just for reference, how much improvemnt was there from the "slow version" to the "fast" version of the IML program? The slow version took ____ seconds; the fast one takes 4 seconds.

Rick_SAS · ‎03-06-2012

Is the data so large that the CHAR, MKT, RET, and STOCKS variables don't fit into memory? If they do fit, then read those variables once, include them as global variables to the opt() module, and index into them inside the DO loop over dates.

Rick_SAS · ‎03-06-2012

One reason that your IML code is slow is that you are re-reading the wgtfracsas and equalsas data sets at every iteration of the optimization, even though those values are unchanged during the optimization.

Rick_SAS · ‎03-06-2012

This isn't really my area, but if you have a stationary process you can use SAS/QC control charts to monitor "normal" versus "abnormal" variation. The SHEWHART procedure has a wide variety of charts for monitoring processes. For example, see these two examples: http://support.sas.com/documentation/cdl/en/qcug/63964/HTML/default/viewer.htm#qcug_shewhart_a0000003904.htm http://support.sas.com/documentation/cdl/en/qcug/63964/HTML/default/viewer.htm#qcug_shewhart_a0000003932.htm There is also the MACONTROL procedure for creating moving average control charts http://support.sas.com/documentation/cdl/en/qcug/63964/HTML/default/viewer.htm#qcug_macontrol_a0000000333.htm

Rick_SAS · ‎03-05-2012

See this blog post: http://blogs.sas.com/content/graphicallyspeaking/2012/02/01/forest-plot-using-sgplot-procedure/

Rick_SAS · ‎03-05-2012

It's not just the explanatory variables, but any variables listed in the CLASS stmt, the MODEL stmt, or the WEIGHT statement. To count the number of missing values in your variables, see this article: http://blogs.sas.com/content/iml/2011/09/19/count-the-number-of-missing-values-for-each-variable/

Rick_SAS · ‎03-05-2012

A more general question is "is a bar chart with a confidence interval a good way to display these data." A discussion of this question is available at http://blogs.sas.com/content/iml/2011/10/07/creating-bar-charts-with-confidence-intervals/

Rick_SAS · ‎03-05-2012

You can contact technical support to see if they have any advice on the out-of-memory situation. Regarding macros, SAS/IML Studio programs are not parsed by the macro facility, so it doesn't support macros in the usual way. However, you can 1) call macros within a SUBMIT/ENDSUBMIT block. These are executed in the SAS environment. 2) Use the SYMGET and SYMPUT family of functions to get or set macro variables from within a SAS/IML Studio program. For example, the following statements set a macro variable (The '@' operator is shorthand for 'submit this single SAS statement.'), then reads in the macro and assigns it to an IML variable: @%let g=5; /* submit global stmt */ five = symgetn("g"); print five; See p. 99 of my book for more examples.

Rick_SAS · ‎03-05-2012

You have several options. Since your response variable is ordinal, read this example to see if ordinal logistic regression fits your needs: http://support.sas.com/documentation/cdl/en/statug/63962/HTML/default/viewer.htm#statug_logistic_sect061.htm

Rick_SAS · ‎03-05-2012

What you say "should work" does work. proc factor data=correl score NFACT=15 outstat=fact; run; proc score data=raw score=fact out=scores; run; If you use NFACT=15, the FACT data set contains scoring coefficients for the first 15 factors. PROC SCORE projects the raw data onto the first 15 factors. The variables are FACTOR1-FACTOR15. Those variables are the first 15 principal components: the linear combinations that best explain the variation in the data.

Rick_SAS · ‎03-05-2012

Strange. SAS didn't change, but you are getting different behavior. I have no idea what has changed on your system, but remember that memory is shared among all applications. If you have a big PDF file open, an email program running, or other applications, those are all using memory, too. However, maybe I can solve the immediate problem of getting your program to run. Instead of dwn=choose(wn=.,0,wn); try this: dwn = wn; /* copy wn */ idx = loc(dwn=.); /* find missing */ if ncol(idx)>0 then dwn[idx]=0; /* replace with zeros */ The above code can be even more memory-efficient if you are willing to overwrite the original copy of wn: idx = loc(wn=.); if ncol(idx)>0 then wn[idx]=0; /* overwrite wn */ Of course, this might just move the out-of-memory error to another line of the program.

Rick_SAS · ‎03-01-2012

The SAS doc has a comparison of the two methods: http://support.sas.com/documentation/cdl/en/proc/63079/HTML/default/viewer.htm#p0gpxoqczfzcyln13ma53kpq38lo.htm

Rick_SAS · ‎03-01-2012

I don't think there are any numerical/statistical differences. I find the CLASS statement convenient when I want to see all of the output in a single table; the BY group approach puts each BY group statistics on a separate page. Also, you need to SORT the data to use the BY group, but not to use the CLASS stmt. The BY group approach is more efficient when the data are sorted, and requires less memory. The output data sets also look different for the two approaches.

Rick_SAS · ‎02-27-2012

I was mistaken when I originally said you get different answers if you turn on ODS GRAPHICS. It's actually asking for the ROCPLOT that causes the different behavior. I have inserted the clause "and ask for PLOTS(ONLY)=ROCPLOT" into my original answer. I don't know of any other procedures that give different answers when a graph is requested, but I didn't know about this one until today. You'll can check with Technical Support if you want an "official" answer to that question. My answers are all unofficial and shouldn't be given any more weight than you'd give to anyone else.

Rick_SAS · ‎02-27-2012

By the way, I determined which PROC run was correct by outputting the ROC curve by using the OUTROC= option and then using the trapezoidal rule to integrate the AUC. You can read about this process in my article "A statistical application of numerical integration: The area under an ROC curve." You can also use SAS/IML to compute the ROC curve from basic principles. It's a good way to understand what PROC LOGISTIC is doing "under the hood."

Online Status	Offline
Date Last Visited	7 hours ago