Missing output from Margins Macro

[closetcoer]

Hi - hope this is the right location so here goes.

I am interested in calculating the predicted probabilities for a binary-level outcome and using a categorical predictor (4-levels from 0-3). 

The Margins Macro appears to have what I need but, when I run the macro, the output does not include a Predictive Margins Table and there is a note that the "algorithm converged". Also no errors noted in the Log file.

Any ideas on what I'm doing wrong? 

Program and output below.

Many thanks.

PROGRAM:

%Margins (data = ecg.ECG_only_final,

            response = lad_axis,*BINARY VARIABLE WITH VALUES OF 1, 2;

            roptions = events='1',

            class = bp_htn,*MULTI-LEVEL VARIABLE WITH VALUES OF 0,1,2,3;

            model = bp_htn,

            dist = binomial,

            margins = bp_htn,

            options = diff cl);

run;

The GENMOD Procedure

Model Information
Data Set	WORK._EXPDATA
Distribution	Binomial
Link Function	Logit
Dependent Variable	LAD_axis	LAD_axis

Number of Observations Read	729
Number of Observations Used	729
Number of Events	38
Number of Trials	729

Response Profile
Ordered Value	LAD_axis	Total Frequency
1	1	38
2	0	691

PROC GENMOD is modeling the probability that LAD_axis='1'. One way to change this to model the probability that LAD_axis='0' is to specify the DESCENDING option in the PROC statement.

Criteria For Assessing Goodness Of Fit
Criterion	DF	Value	Value/DF
Log Likelihood		-147.6443
Full Log Likelihood		-147.6443
AIC (smaller is better)		303.2885
AICC (smaller is better)		303.3438
BIC (smaller is better)		321.6552

Algorithm converged.

Analysis Of Maximum Likelihood Parameter Estimates
Parameter	DF	Estimate	Standard Error	Wald 95% Confidence Limits	Wald Chi-Square	Pr > ChiSq
Intercept	1	-2.6741	0.2585	-3.1808	-2.1675	107.04	<.0001
bp_htn_0	1	-0.6217	0.4129	-1.4310	0.1876	2.27	0.1322
bp_htn_1	1	0.0351	0.4171	-0.7824	0.8525	0.01	0.9329
bp_htn_2	1	-0.5245	0.7663	-2.0264	0.9774	0.47	0.4937
bp_htn_3	0	0.0000	0.0000	0.0000	0.0000	.	.
Scale	0	1.0000	0.0000	1.0000	1.0000

Note:

The scale parameter was held fixed.

[unison · Accepted solution]

What output tables are you getting? According to this (http://support.sas.com/kb/63/038.html#use), the margins should be in and output table bp_htn_margins.

-unison

[closetcoer]

Thanks so much unison - I tried rebooting my system (long story) and then reran the macro. The expected tables showed up as designed so, while I'm not at all sure what the issue was, the reboot seems to have done the trick. I appreciate your time.

[Reeza]

Did you run it on their sample data to test it first? Did that generate the tables you want?

[closetcoer]

Thanks Reeza - see my reply to Unison above. After a system reboot and rerunning the macro, I got what I expected in terms of output tables. FYI, there was no sample data file but I think one can be created with the frequency information in the macro documentation. Thanks for yoru time.

[Reeza]

---

@closetcoer wrote:
FYI, there was no sample data file but I think one can be created with the frequency information in the macro documentation. Thanks for yoru time.

---

http://support.sas.com/kb/63/038.html#use

They provide the data to run the examples, which should be a first step when testing code/macros.

EXAMPLE 1: Predictive margins in a binary logistic modelThe following statements estimate and test predictive margins for the Neuralgia data set presented in the example titled "Logistic Modeling with Categorical Predictors" in the LOGISTIC documentation. Treatment predictive margins are computed for males and females. Since the Margins macro requires a numeric response variable, the NoPain variable is created with value 1 when Pain='No' and 0 otherwise. The probability of no pain is modeled as a result of roptions=event='1'. Confidence intervals are requested by options=cl.

data Neur; 
  set Neuralgia; 
  NoPain=(pain='No'); 
  run;
%Margins(data     = Neur,
         class    = Treatment Sex,
         response = NoPain,
         roptions = event='1',
         dist     = binomial,
         model    = Treatment Sex Treatment*Sex Age Duration,
         margins  = Treatment, 
         at       = Sex,
         options  = cl )