overlay fitted density Poisson data

[Ismene]

Dear SAS Help,

I use SAS 9.4

I tried the exercise on this link: https://blogs.sas.com/content/iml/2012/04/04/fitting-a-poisson-distribution-to-data-in-sas.html

This is the code I copied from the site.

--------------------------------------

data MyData;
input N @@;
datalines;
7 7 13 9 8 8 9 9 5 6 6 9 5 10 4 5 3 8 4
;
run;

proc freq data=MyData;
tables N / out=FreqOut plots=FreqPlot (scale=percent);
run;

proc genmod data=MyData;
model N= /dist=poisson;
output out= PoissonFit p=lambda;
run;

data _null_;
set Poissonfit;
call symputx("Lambda", Lambda);
stop;
run;

data PMF;
do t= 0 to 13;
Y = pdf("Poisson", t, &Lambda);
end;
run;

data Discrete;
merge FreqOut PMF;
Prop = Percent / 100;
run;


*\bij dit stukje is er een probleem. Ik heb gemaild naar de site op 15 januari;
proc sgplot data=Discrete;
vbarparm category=N response=Prop /legendlabel = 'Sample';
scatter x=T y=Y / legendlabel= 'PMF'
markerattrs=GraphDataDefault(symbol=CIRCLEFILLED size=10);
title "Emails per 30minute Poissondist";
run;

-----------------------------------------------------------------------

The log gives this error message:

NOTE: Since no format is assigned, the numeric category variable will use the default of BEST6.

The plot that I should get should look like the upper plot, but my result is the lower plot. 

Questions:

1. What did I do wrong?

2. Is it possible to have a line instead of dots.

3. Can I use these steps to do an overlay on a poisson density plot created with the FMM procedure?

Thanks in advance for your answer.

The plot I get looks like this:

[ABritinAus]

Hi

The dataset PMF probably doesn't contain what you expect.

Try adding the word output as below.  Then you should get the observations you expect.

data PMF; do t= 0 to 13;
Y = pdf("Poisson", t, &Lambda); output;
end;
run;

To get a line instead of circles, try changing the word scatter to series.

Re question 3: sorry, I don't know.

Hope that helps.

[PeterClemmensen]

1) You need an OUTPUT Statement in your do look. As of now, you are creating just 1 observation in your PMD data set

data PMF;
do t= 0 to 13;
Y = pdf("Poisson", t, &Lambda);
output;
end;
run;

2) Yes, it is possible, simply change the SCATTER Statement to a SERIES Statement like this. Use the MARKERATTRS Option to control the markers and delete both the MARKER and MARKERATTRS Options 

proc sgplot data=Discrete;
vbarparm category=N response=Prop /legendlabel = 'Sample';
series x=T y=Y / markers legendlabel= 'PMF'
markerattrs=GraphDataDefault(symbol=CIRCLEFILLED size=10);
title "Emails per 30minute Poissondist";
run;

3) Show us your PROC FMM code 🙂 You can always extract underlying data from procedure output with an ODS OUTPUT statement.

[Ismene]

Thank you very much. This worked.

Here is my proc fmm code. The data file is attached.

proc fmm data=acu plots=density (bins=15);
model frequency= /dist=poisson  k=3 
parms(2.1, 2.8, 3.2);  *\starting values for mean ln(8), ln(16) en ln(24); 
probmodel / parms (0, 0); 
freq age;
title 'poisson with 3 support points';
run;

[PeterClemmensen]

Do something like this and change the Component value in the Where Statement to get another Component model 🙂

proc fmm data=acu plots=density (bins=15);
model frequency= /dist=poisson  k=3 
parms(2.1, 2.8, 3.2);  *\starting values for mean ln(8), ln(16) en ln(24); 
probmodel / parms (0, 0); 
freq age;
title 'poisson with 3 support points';

ods output ParameterEstimates=pePoisson;
run;

data _null_;
	set pePoisson;
	where Component=1;
	call symputx('lambdaFMM', Estimate);
run;

%put &lambdaFMM.;

data PMF_fmm;
do t= 0 to 13;
Y = pdf("Poisson", t, &lambdaFMM.);
output;
end;
run;

data Discrete_fmm;
merge FreqOut PMF_fmm;
Prop = Percent / 100;
run;

proc sgplot data=Discrete_fmm;
vbarparm category=N response=Prop /legendlabel = 'Sample';
series x=T y=Y / markers legendlabel= 'PMF'
markerattrs=GraphDataDefault(symbol=CIRCLEFILLED size=10);
title "Emails per 30minute Poissondist";
run;