cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Search instead for 
Did you mean: 
mgrasmussen
Monday
  • 87 Posts
  • 54 Likes Given
  • 0 Solutions
  • 4 Likes Received

How to include a number immediately after one macro and immediately be...

by Quartz | Level 8 in SAS Programming
a week ago
a week ago
Dear SAS experts   I have made a simplified example of the problem I am facing.   Following the code in the bottom SAS reports:   SYMBOLGEN: Macro variable J resolves to 2008 SYMBOLGEN: Macro variable L resolves to 3 bef20083 SYMBOLGEN: Macro variable J resolves to 2008 SYMBOLGEN: Macro variable L resolves to 6 bef20086 SYMBOLGEN: Macro variable J resolves to 2008 SYMBOLGEN: Macro variable L resolves to 9 bef20089 SYMBOLGEN: Macro variable J resolves to 2008 SYMBOLGEN: Macro variable L resolves to 12 bef200812 SYMBOLGEN: Macro variable J resolves to 2009 SYMBOLGEN: Macro variable L resolves to 3 bef20093 SYMBOLGEN: Macro variable J resolves to 2009 SYMBOLGEN: Macro variable L resolves to 6 bef20096 SYMBOLGEN: Macro variable J resolves to 2009 SYMBOLGEN: Macro variable L resolves to 9 bef20099 SYMBOLGEN: Macro variable J resolves to 2009 SYMBOLGEN: Macro variable L resolves to 12   I would like to include a 0 in front of 3, 6 and 9, but it is not included based on my code. Can anyone help me resolve this issue.   Thank you   Here is the code:   %macro test;   %do j=2008 %to 2009 %by 1;             %do l=3 %to 12 %by 3;                       %if l<12 %then %do;                       %put bef&j.0&l;                        %end;                        %else %do;                        %put bef&j.&l;                         %end;             %end;   %end;   %mend;   %test; ... View more

Re: Potential issue with lsmeans in proc mixed (output: Non-est)

by Quartz | Level 8 in Statistical Procedures
‎03-13-2023 07:10 AM
‎03-13-2023 07:10 AM
Dear Steve   Thank you very much for your reply.    To be frank I do not have a straight answer to your question. I am a non-statistician who tried to build a model (only with brief statistical consult), but I realize now that I might not fully understand the working of the splines.   In simple terms what I want to do is to visually illustrate and statistically test for the differences in slopes in hba1c for men and women. More specifically, I want to investigate whether there is a difference in mean hba1c over time (time on the x-axis) between men and women. However, I also want to be able to test means differences at specific time points.   My understanding was that the four spline variables all-in-all represented time and thus that I had to model an interaction term for each splines variable. I sort of "transferred" the logic I would apply to a normal OLS regression to this new model. In OLS - here assumping a linear relationship between days/time and hba1c - I would write the following to test for interaction of sex in the relationship between time and hba1c:   proc reg data=datasetname; class sex; model hba1c=sex days sex*days; run;   For the time being disregard the fact that I am actually using a mixed effect model. The point here is merely an attempt to try to explain the background for how I am applying the splines in the analysis. However incorrectly.    I realize that I may not be using the correct statistical jargon.    I would be happy to receive any feedback and suggestion on moving forward.   I mentioned earlier, when I only include the first spline variable I don't experience an issue with computing LSmeans. But I am not sure about the significance of this.   Thank you  ... View more

Re: Potential issue with lsmeans in proc mixed (output: Non-est)

by Quartz | Level 8 in Statistical Procedures
‎03-09-2023 02:38 AM
‎03-09-2023 02:38 AM
Dear Jill   Thank you for the feedback.   I will bring all of the feedback I have gotten, including yours most recently, back to my group and we will discuss the analyses.   Best regards   Martin ... View more

Re: Potential issue with lsmeans in proc mixed (output: Non-est)

by Quartz | Level 8 in Statistical Procedures
‎03-07-2023 05:30 AM
‎03-07-2023 05:30 AM
Right, it is an issue about whether or not both males and females are represented at each level of spl1-spl4?   Thanks for the suggested syntax. I ran it and there are many places with empty cells/cell with 0 observations and this is true for all four spline variables. This is also the case for spl1 and Sex. As mentioned when I include the interaction between the two in the model I don't get the "Non-est". This particular model appears to work. ... View more

Re: Potential issue with lsmeans in proc mixed (output: Non-est)

by Quartz | Level 8 in Statistical Procedures
‎03-07-2023 03:37 AM
‎03-07-2023 03:37 AM
Dear PaigeMiller   Thank you very much for your reply.   "Often, a Least Squares Mean is non-estimable because you have cells in your experiment where there is no data." Missing data in general is not an issue in this dataset, but their might be a specific issue relating to the interaction terms. Is there some simple way to check for this? As far as I see it, there are four spline variables and the first spline variable corresponds to the variable Days - from which the spline variables are derived - in a 1:1 fashion. As mentioned, if I create a model that includes only the interaction spl1*sex (and thus I do not include spl2*sex, spl3*sex and spl4*sex) I don't get "Non-est". But only if I remove all three. Removing just one of them does not change the outcome.   "And as said by @StatsMan , the model you have chosen to fit "might be more than you can do with this data." I realize that there might be an issue here. When I completely remove the random statement I still get the "Non-est" outcome, which suggest to me that there is an issue with the abovementioned interaction terms.   Thank you   Best regards ... View more

Re: Potential issue with lsmeans in proc mixed (output: Non-est)

by Quartz | Level 8 in Statistical Procedures
‎03-07-2023 03:27 AM
‎03-07-2023 03:27 AM
Dear Statsman    I much appreciate your reply. 1. "You can let GLIMMIX do the dirty work of forming the spline basis using the EFFECT statement as in this example"   I am not sure exactly what you mean by this. The EFFECT statement is not supported by Proc Mixed. Moreover, would it make a difference if it was? On another note I am actually using the EFFECT statement in proc glmselect to create the spline basis:   proc glmselect data=dataset_withoutsplines outdesign(addinputvars fullmodel)=dataset_withsplines;  effect spl = spline(hba1c / naturalcubic basis=tpf(noint) knotmethod=percentiles(5)); model hba1c = spl / selection=none;  quit;   2. a) "Fitting random coefficients on top of the spline fit might be more than you can do with this data." and b) "You could try fitting just the random intercept and see if that does better for you." Right, I understand that the model might be too complex given the data. Regarding 2b), even when I completely remove the random statement I get the same outcome of "Non-est". This suggest to be that there is an issue with the interactions of sex and spl2-spl4. As I mentioned in my original post I don't get "Non-est" when I only include the interaction spl1*sex. However, I do not know if this is an appropriate model.   3. "If you do want those random coefficients for each subject, do you want those on the spline effect rather than a linear effect of days?" Are you suggesting that I could include all the spline variables here? random intercept SPLINE VARIABLES / subject = pseudo_id type=un;   I have not considered this. Could one argue for and against doing such?   Best regards ... View more

Potential issue with lsmeans in proc mixed (output: Non-est)

by Quartz | Level 8 in Statistical Procedures
‎03-03-2023 07:07 AM
‎03-03-2023 07:07 AM
Dear SAS experts   I am a non-statistician who is seeking some help with getting basic syntax running.   I am trying to run a model using proc mixed, but I appear to be having an issue when using lsmeans.   First, I have created restricted cubic basis functions (4 new variables; spl1-spl4) based on the variable days. Days is a continuous variable that includes days since 1 st of January 2020 to 31 st of December 2021.   For each value of days, there is a measure of Hba1c. Hba1c is a clinical measure of blood sugar levels during the last three months. It varies in values from 7-173.   To begin with (just to get the basic syntax working), I want to model the splines of days and the variable sex and estimate the general between-sex difference in mean hba1c. I am using a mixed-effect model to take into account that there are multiple measurements of hba1c for each individual (the amount varies between subjects).   I have written the following syntax in proc mixed:   proc mixed data=dataset_name; class pseudo_id sex; model hba1c=spl: sex sex*spl1 sex*spl2 sex*spl3 sex*spl4 / solution;            random intercept days / subject = pseudo_id type=un; lsmeans sex / pdiff; run;   The log output is: 1          ;*';*";*/;quit;run; 2          OPTIONS PAGENO=MIN; 3          %LET _CLIENTTASKLABEL='7. Hba1c_analysis'; 4          %LET _CLIENTPROCESSFLOWNAME='Process Flow'; 5          %LET _CLIENTPROJECTPATH=''; 6          %LET _CLIENTPROJECTPATHHOST=''; 7          %LET _CLIENTPROJECTNAME=''; 8          %LET _SASPROGRAMFILE='T:\ouh\afd\covid_forskningsartikel\Programfiler og output\7. Hba1c_analysis.sas'; 9          %LET _SASPROGRAMFILEHOST='L111353'; 10         11         ODS _ALL_ CLOSE; 12         OPTIONS DEV=PNG; 13         GOPTIONS XPIXELS=0 YPIXELS=0; 14         FILENAME EGSR TEMP; 15         ODS tagsets.sasreport13(ID=EGSR) FILE=EGSR 16             STYLE=HtmlBlue 17             STYLESHEET=(URL="file:///C:/Program%20Files/SASHome/SASEnterpriseGuide/7.1/Styles/HtmlBlue.css") 18             NOGTITLE 19             NOGFOOTNOTE 20             GPATH=&sasworklocation SYMBOLGEN:  Macro variable SASWORKLOCATION resolves to "E:\Work\_TD2300848_SRVESBAPPSAS34V_\Prc2/" 21             ENCODING=UTF8 22             options(rolap="on") 23         ; NOTE: Writing TAGSETS.SASREPORT13(EGSR) Body file: EGSR 24         25         GOPTIONS ACCESSIBLE; 26         proc mixed data=hba1c_sdco6; 27         class pseudo_id køn_20; 28         model visres_numerisk=spl: køn_20 køn_20*spl1 køn_20*spl2 køn_20*spl3 køn_20*spl4 / solution outp=pred6r outpm = pred6f;         29         random intercept prvtdato_num / subject = pseudo_id type=un; 30         lsmeans køn_20 / pdiff at means; 31         run;   NOTE: 203 observations are not included because of missing values. NOTE: Convergence criteria met. MVA_DSIO.OPEN_CLOSE| _DISARM|         STOP| _DISARM| 2023-03-03T11:49:59,790+01:00| _DISARM| WorkspaceServer| _DISARM| | _DISARM| | _DISARM| | _DISARM| 28151808| _DISARM| 11| _DISARM| 18| _DISARM| 194607184| _DISARM| 3544470331| _DISARM| | _DISARM| | _DISARM| | _DISARM| | _DISARM| | _DISARM| | _ENDDISARM NOTE: The data set WORK.PRED6R has 22721 observations and 26 variables. NOTE: Compressing data set WORK.PRED6R decreased size by 27.40 percent.       Compressed is 53 pages; un-compressed would require 73 pages. MVA_DSIO.OPEN_CLOSE| _DISARM|         STOP| _DISARM| 2023-03-03T11:49:59,790+01:00| _DISARM| WorkspaceServer| _DISARM| | _DISARM| | _DISARM| | _DISARM| 28151808| _DISARM| 11| _DISARM| 18| _DISARM| 194936189| _DISARM| 3544799946| _DISARM| | _DISARM| | _DISARM| | _DISARM| | _DISARM| | _DISARM| | _ENDDISARM NOTE: The data set WORK.PRED6F has 22721 observations and 26 variables. NOTE: Compressing data set WORK.PRED6F decreased size by 27.40 percent.       Compressed is 53 pages; un-compressed would require 73 pages. MVA_DSIO.OPEN_CLOSE| _DISARM|         STOP| _DISARM| 2023-03-03T11:49:59,790+01:00| _DISARM| WorkspaceServer| _DISARM| | _DISARM| | _DISARM| | _DISARM| 28151808| _DISARM| 11| _DISARM| 18| _DISARM| 195132996| _DISARM| 3544997465| _DISARM| | _DISARM| | _DISARM| | _DISARM| | _DISARM| | _DISARM| | _ENDDISARM PROCEDURE| _DISARM|         STOP| _DISARM| 2023-03-03T11:49:59,790+01:00| _DISARM| WorkspaceServer| _DISARM| | _DISARM| | _DISARM| | _DISARM| 28151808| _DISARM| 11| _DISARM| 18| _DISARM| 195398608| _DISARM| 3544999001| _DISARM| | _DISARM| | _DISARM| | _DISARM| | _DISARM| | _DISARM| | _ENDDISARM NOTE: PROCEDURE MIXED used (Total process time):       real time           55.48 seconds       cpu time            55.43 seconds       2                                                          The SAS System                                10:41 Friday, March 3, 2023     32         33         GOPTIONS NOACCESSIBLE; 34         %LET _CLIENTTASKLABEL=; 35         %LET _CLIENTPROCESSFLOWNAME=; 36         %LET _CLIENTPROJECTPATH=; 37         %LET _CLIENTPROJECTPATHHOST=; 38         %LET _CLIENTPROJECTNAME=; 39         %LET _SASPROGRAMFILE=; 40         %LET _SASPROGRAMFILEHOST=; 41         42         ;*';*";*/;quit;run; 43         ODS _ALL_ CLOSE; 44         45         46         QUIT; RUN; 47             The results are: Dimensions Covariance Parameters 4 Columns in X 15 Columns in Z per Subject 2 Subjects 3434 Max Obs per Subject 27   Number of Observations Number of Observations Read 22721 Number of Observations Used 22518 Number of Observations Not Used 203   Iteration History Iteration Evaluations -2 Res Log Like Criterion 0 1 178669.59663948   1 3 161030.12170074 0.00065048 2 1 160985.21896392 0.00005265 3 1 160981.89111021 0.00000044 4 1 160981.86465805 0.00000000   Convergence criteria met.   Covariance Parameter Estimates Cov Parm Subject Estimate UN(1,1) pseudo_id 137.38 UN(2,1) pseudo_id -0.07511 UN(2,2) pseudo_id 0.000190 Residual   42.5146   Fit Statistics -2 Res Log Likelihood 160981.9 AIC (Smaller is Better) 160989.9 AICC (Smaller is Better) 160989.9 BIC (Smaller is Better) 161014.4   Null Model Likelihood Ratio Test DF Chi-Square Pr > ChiSq 3 17687.73 <.0001   Solution for Fixed Effects Effect køn_20 Estimate Standard Error DF t Value Pr > |t| Intercept   61.0501 0.3376 3432 180.82 <.0001 spl1   -0.01390 0.001344 16E3 -10.34 <.0001 spl2   0.000203 0.000021 16E3 9.66 <.0001 spl3   -0.00050 0.000054 16E3 -9.33 <.0001 spl4   0.000442 0.000050 16E3 8.78 <.0001 Sex Female 0.2017 0.5318 16E3 0.38 0.7045 Sex Male 0 . . . . spl1*Sex Female 0.01093 0.002127 16E3 5.14 <.0001 spl1*Sex Male 0 . . . . spl2*Sex Female -0.00019 0.000033 16E3 -5.56 <.0001 spl2*Sex Male 0 . . . . spl3*Sex Female 0.000450 0.000085 16E3 5.26 <.0001 spl3*Sex Male 0 . . . . spl4*Sex Female -0.00039 0.000080 16E3 -4.87 <.0001 spl4*Sex Male 0 . . . .   Type 3 Tests of Fixed Effects Effect Num DF Den DF F Value Pr > F spl1 1 16E3 62.99 <.0001 spl2 1 16E3 93.40 <.0001 spl3 1 16E3 87.02 <.0001 spl4 1 16E3 77.01 <.0001 Sex 1 16E3 0.14 0.7045 spl1*Sex 1 16E3 26.40 <.0001 spl2*Sex 1 16E3 30.88 <.0001 spl3*Sex 1 16E3 27.70 <.0001 spl4*Sex 1 16E3 23.68 <.0001   Least Squares Means Effect Sex spl1 spl2 spl3 spl4 prvtdato_num Estimate Standard Error DF t Value Pr > |t| Sex Female 407.07 105217 52728 19728 407.07 60.1981 0.2895 16E3 207.93 <.0001 Sex Male 407.07 105217 52728 19728 407.07 Non-est . . . .   Differences of Least Squares Means Effect Sex _Sex spl1 spl2 spl3 spl4 prvtdato_num Estimate Standard Error DF t Value Pr > |t| Sex Female Male 407.07 105217 52728 19728 407.07 Non-est . . . .   As you can see there is no estimate for males, and thus the differences between the sexes cannot be estimated. Lsmeans produces an output for males when I only include the interaction: Sex*spl1 (but not for the other interactions).   Can anyone help me solve this issue?   Thank you ... View more

Re: Defining a table title based on macro parameter

by Quartz | Level 8 in SAS Programming
‎02-22-2022 09:08 AM
‎02-22-2022 09:08 AM
Dear PaigeMiller    Thanks.   I was not aware of the fact that I also needed a "%end" there. So, what was lacking was because of ignorance.   Regarding your second point, I want to include some additional text and not just the "&Sex." part. In the actual data what I want to specificy is that the statistics I am summarizing are based on different age group restrictions depending on the (disease) subgroup of the population. Therefore, I would like to be able to some how write this out. ... View more

Defining a table title based on macro parameter

by Quartz | Level 8 in SAS Programming
‎02-22-2022 08:54 AM
‎02-22-2022 08:54 AM
Dear SAS experts   I would like to create a table where the title is defined depending on the macro parameter specified. Below is a simple example of scenario. If "M" is specified I would like a title specific for "M", and the same for "F". I cannot get it to work, but I suspect I need to somehow use %if and %else in the macro. Can someone help with the code? Thanks.   data mydata; set sashelp.class; run;   *Creating tables; %macro table(Sex=); proc means data=mydata; var Age; where Sex="&Sex."; %if &Sex.="M" %then %do; Title "Table of: &Sex. (M text)"; %else %do; Title "Table of: &Sex. (F text)"; %end; run;   %mend;   %table(Sex=M); %table(Sex=F); ... View more

Re: Change default coloring in stacked bar chart? (sgplot, hbar..)

by Quartz | Level 8 in SAS Programming
‎02-07-2022 06:31 AM
‎02-07-2022 06:31 AM
I think the simple version will work for now. Thanks. ... View more

Re: Change default coloring in stacked bar chart? (sgplot, hbar..)

by Quartz | Level 8 in SAS Programming
‎02-07-2022 06:10 AM
‎02-07-2022 06:10 AM
Thanks a bunch, Rick. This looks promising.   The variable in group= is not constant. The basic difference between the (many) graphs is the variable that I include in group, but more so the population and year that the graph represents.   So, let's say I have 20 different sgplot with each their own specific variable in group=. Some of the variables will have 2 categories and some will have 5 or more. I might make something like 32 different charts (different years, different populations) based on each sgplot code. I don't mind making a specification for styleattrs datacolors for each of the "main" 20 different sgplot codes. Actually, there might be some benefit to doing that. If you see some issue with this I will be happy to hear about it. ... View more

Change default coloring in stacked bar chart? (sgplot, hbar..)

by Quartz | Level 8 in SAS Programming
‎02-07-2022 05:32 AM
‎02-07-2022 05:32 AM
Dear SAS programmers   I am creating a lot of stacked bar charts in SAS using sgplot. I have created an example of the type of graph I am creating;   data have; set sashelp.cars; run; proc sgplot data=have; hbar make/stat=percent group=DriveTrain seglabel; yaxis type=discrete fitpolicy=none; run;   I am relatively new to SAS. I observe that there is some sort of default or hiarchical coloring of the stacked charts; blue, red, green, some yellow-ish color etc.    Is there some simple way to change this default coloring, fx to different shades of green?   I have seen other SAS users ask similar questions. Someone suggested that you create a dataset with certain color information and then use attrmap= and attrid= in the sgplot code. However, is there not a simple way of simple stating: use these colors and in this order? My issue is that I would like to create many graphs and not deal with each "bar-variable".   Thank you ... View more

Re: Get code to work in an array

by Quartz | Level 8 in SAS Programming
‎01-07-2022 07:38 AM
‎01-07-2022 07:38 AM
By "dimension" do you mean number of elements in the array? So the "8" says something about how large the numbers included in the array retains are expected to be? I am relatively new to SAS coding.   I slightly changed some of your code. I removed the the by-code and the globalcatvar, which I thought was neccessary. The code seems to be doing what I was hoping for. Moreover, it does appear that the code below would work without any knowledge about the number of numeric variables in the dataset?   data have; input varone vartwo varthree varfour; datalines; 1.1 2 1.5 5 1 2 1 1 1 3 1 2 1 4 3 10 ; run;   data want_three; set have; array modify {*} _NUMERIC_; array retains (200) 8 _temporary_; if _n_=1 then do i=1 to dim(modify); retains(i)=modify(i); end; else do i=1 to dim(modify); modify{i}=modify{i}/retains(i); end; run; ... View more

Re: Get code to work in an array

by Quartz | Level 8 in SAS Programming
‎01-07-2022 07:27 AM
‎01-07-2022 07:27 AM
Hey Andreas   Thanks. No, I might not need globalcatvar.   Your code looks interesting. Is it possible to modify it such that all numeric variables are included in vars (and that the rest of the code can be modified to work too)?   Thanks ... View more
Contact Me
Online Status
Offline
Date Last Visited
Monday
Likes given to
Likes from
View All
Quick links