Proc Surverymeans Unexpected Note omitting stratum

[rkobilis]

Greetings all. I am working on a project in which immunization data from Childcare Group Settings is analyzed. The SAS Program that is used to run the analysis utilizes Proc Surveymeans.

The Strata are 8 different Districts in my state.

The Clusters are the individual Childcare Settings.

And the Domain are 5 different Age Groups (2-3 months, 4-5 months, 6-18 months, 19-35 months, and 36-59 months).

I am receiving a Note in the Log which appears to have no Logic behind it:

"NOTE: Only one cluster in a stratum for Age_Group_Num domains for variable(s) Up_to_Date, DTap!, DTaP2, Polio1, Polio2, Hib1, Hib2, HepB1, HepB2, PCV1, PCV2, Rota1, Rota2, exemption_cat. The estimate of variance for Up_to_Date, DTap!, DTaP2, Polio1, Polio2, Hib1, Hib2, HepB1, HepB2, PCV1, PCV2, Rota1, Rota2, exemption_cat will omit this stratum."

To me this note implies that in a particular District(Strata), that there is only one Childcare Setting(Cluster) that serves a particular Age Group(Domain). But this is not the case. There are at least 2 Clusters that serve each of the Age Groups(Domain). Thus I do not understand why this note is presenting.

Have I misinterpreted the Note? In trial and error I noticed that when the Strata Statement is removed from the Proc that the Note no longer presents, and the confidence intervals are smaller (which is usually good, but accuracy is more important).

Any help into this matter would be greatly appreciated!!

Thank you.

[Reeza]

Are any of the variables in the list above missing for that stratum? If so, that row would be excluded from analysis.

[ballardw]

It is usually a good idea to post the code and messages from the log. Copy from the log and paste into a code box opened using the forum's {I} menu icon to preserve formatting which is important in the case of error messages as the main message windows will reformat text and move some of the characters indicating locations of errors.

The code will help us see if perhaps your actual analysis is asking for more (or less) than your narrative.

It also lets us see actual variable names.

I would be tempted to run code that looks something like

proc freq data=have;
   tables district*caresetting*agegroup /list nopercent missing;
run;

to check most of the culprits. Note that without code I am providing functional descriptions instead of your actual variable or dataset names...

[rkobilis]

/*Statewide*/
proc surveymeans data = Results3_PA total = InputPSU_Totals;

/*Lists variables that are to be analyzed as categorical.*/
class	Up_to_Date Series_431331 Series_43133 Series_431
		DTaP1 DTaP2 DTaP3 DTaP4 Polio1 Polio2 Polio3 Hib1 Hib2 Hib3 Hib4 HepB1 HepB2 HepB3 MMR1 Varicella1 PCV1 PCV2 PCV3 PCV4 Flu1 HepA1 HepA2 Rota1 Rota2 exemption_cat;
/*Lists variables that indicate clusters.*/
cluster	UniqueID;
/*Lists variables that indicate subpopulation analyses required.*/
domain	Age_Group_Num;
/*Lists the variables that indicate the strata.*/
strata	district;
/*Lists variables to be analyzed.*/
var 	Up_to_Date Series_431331 Series_43133 Series_431
		DTaP1 DTaP2 DTaP3 DTaP4 Polio1 Polio2 Polio3 Hib1 Hib2 Hib3 Hib4 HepB1 HepB2 HepB3 MMR1 Varicella1 PCV1 PCV2 PCV3 PCV4 Flu1 HepA1 HepA2 Rota1 Rota2 exemption_cat;

/*Indicates the weight variable.*/
weight	final_weight;

/*Names the output dataset.*/
/*StatsOut_PA1 contains the information for Table 3.  StatsOutAllAge_PA1 contains the information on all ages to be used to report exemption categories.*/
ods output Domain = StatsOut_PA1 Statistics = StatsOutAllAge_PA1;

run;

Hi Ballardw, I apologize for not including the code, datasets, and Log message.

NOTE: Only one cluster in a stratum for Age_Group_Num domains for variable(s) Up_to_Date, DTaP1,
      DTaP2, Polio1, Polio2, Hib1, Hib2, HepB1, HepB2, PCV1, PCV2, Rota1, Rota2, exemption_cat.
      The estimate of variance for Up_to_Date, DTaP1, DTaP2, Polio1, Polio2, Hib1, Hib2, HepB1,
      HepB2, PCV1, PCV2, Rota1, Rota2, exemption_cat will omit this stratum.
NOTE: The data set WORK.STATSOUTALLAGE_PA1 has 62 observations and 7 variables.
NOTE: The data set WORK.STATSOUT_PA1 has 310 observations and 9 variables.
NOTE: PROCEDURE SURVEYMEANS used (Total process time):
      real time           11.72 seconds
      cpu time            11.15 seconds

I have tried attaching both needed Datasets "Results3_PA" And "InputPSU_Totals" but it may be that only 1 attachment may be added per message. If that is the case I will reattach "InputPSU_Totals" as another reply.

[rkobilis]

Yup, looks like only 1 attachment allowed. Here is the other pertinent dataset: