Proc means weight option - explain SAS Help example

[iank131]

Hi,

In SAS 9.2 Help, there is an example entitled "Weighted Statistics Example" that uses the Proc Means Weight option. I am trying to understand how the variances in the example are calculated by doing it manually in EXCEL, but I cannot work out how SAS calculates it. Referring to the code below, I am able to calculate the variance that is in Example 1, but can someone please explain to me Example 2 and 3.

My exact questions are:

1. In example 2, how do you calculate Est_SigmaSq (i.e. var)?

2. If you were to transform the variable ObjectSize in Example 2, what would it be in order to get the variance Est_SigmaSq?

3. The same two questions apply for Example 3.

Below is the code as given in SAS Help and I also attach the Excel I am using to do my manual calculations if you need it.

Thanks and Regards

Ian.

data size;

   input Distance ObjectSize @@;

   Precision=1/distance;

   datalines;

1.5 30 1.5 20 1.5 30 1.5 25

3   43 3   33 3   25 3   30

4.5 25 4.5 36 4.5 48 4.5 33

6   43 6   36 6   23 6   48

7.5 30 7.5 25 7.5 50 7.5 38

;

*EXAMPLE1;

proc means data=size maxdec=3 n mean var stddev;

   var objectsize;

   title1 'Unweighted Analysis of the SIZE Data Set';

run;

**EXAMPLE2;

proc means data=size maxdec=3 n mean var stddev;

   weight precision;

   var objectsize;

   output out=wtstats var=Est_SigmaSq std=Est_Sigma;

   title1 'Weighted Analysis Using Default VARDEF=DF';

run;

**EXAMPLE3;

proc means data=size maxdec=3 n mean var std vardef=weight;

   weight precision;

   var objectsize;

   title1 'Weighted Analysis Using VARDEF=WEIGHT';

run;

[1zmm]

The documentation for the WEIGHT statement in PROC MEANS.describes how the sample variance is calculated from the individual variable values and their weights, the weighted mean, and the divisor[the default value of n-1 or of the sum of the weights when VARDEF=WEIGHT].

[iank131]

1zmm, thanks for answering, but I have read the documentation that you refer to (a number of times) and it does not answer my question. That is why I am asking if someone can help me here.

Ian.

[Reeza]

If you want to mimic the calculation in Excel, take the weight and multiply it by the observation, precision*object_size

Then calculate STD and Variance as you would normally in excel using the weighted observations.

[iank131]

Ok. Thanks. I have been able to reproduce the results.