how to access a summary statistic in the data step?

[eggman2001]

hello,

I know that with proc means I can find summary statistics for my data such as mean and N per variable. However, how can I then use these summary statistics in my data step, so that I can do something like find the distance of each observation from the mean, and then finally add up these distances.

i guess that sometimes i want to operate at the per observation level, and then sometimes i want to operate at the aggregate level, and i'm not quite sure what the approach is to switching back and forth between these. hope i'm making some sense.

[Linlin]

is the example helpful:

proc sql noprint;

  select mean(height) into :mn

   from sashelp.class;

quit;

data class;

   set sashelp.class;

    mean_height=&mn;

  diff=height-&mn;

  proc print;run;

Obs    Name       Sex    Age    Height    Weight     height        diff

        1    Alfred      M      14     69.0      112.5    62.3368      6.6632

        2    Alice       F      13     56.5       84.0    62.3368     -5.8368

        3    Barbara     F      13     65.3       98.0    62.3368      2.9632

        4    Carol       F      14     62.8      102.5    62.3368      0.4632

        5    Henry       M      14     63.5      102.5    62.3368      1.1632

        6    James       M      12     57.3       83.0    62.3368     -5.0368

        7    Jane        F      12     59.8       84.5    62.3368     -2.5368

        8    Janet       F      15     62.5      112.5    62.3368      0.1632

        9    Jeffrey     M      13     62.5       84.0    62.3368      0.1632

       10    John        M      12     59.0       99.5    62.3368     -3.3368

       11    Joyce       F      11     51.3       50.5    62.3368    -11.0368

       12    Judy        F      14     64.3       90.0    62.3368      1.9632

       13    Louise      F      12     56.3       77.0    62.3368     -6.0368

       14    Mary        F      15     66.5      112.0    62.3368      4.1632

       15    Philip      M      16     72.0      150.0    62.3368      9.6632

       16    Robert      M      12     64.8      128.0    62.3368      2.4632

       17    Ronald      M      15     67.0      133.0    62.3368      4.6632

       18    Thomas      M      11     57.5       85.0    62.3368     -4.8368

       19    William     M      15     66.5      112.0    62.3368      4.1632

[Reeza]

Add the statistics back to your datastep. Search on the forum for many ways to do that. This is especially useful if you have statistics at a group level.

You can also look at some of the other stats that proc means gives you because they can be useful.

[Ksharp]

Or something like :

proc means;

..

output out=stat  .....;

run;

data want;

set have;

if _n_ eq 1 then set stat ;

...........

Ksharp

[Prit]

You can add ods statement to collect statistics into new dataset which might be helpful to you.

  ODS OUTPUT summary=summary_means; /* summary_means is the new dataset */

  PROC MEANS DATA= <DATASET_NAME>;

    VAR years_on_farm;

  RUN;

  ODS OUTPUT CLOSE;

[cau83]

Instead of MEANS, use SQL to calculate statistics and then load them into macro variables that you reference in the data step.

eg.

PROC SQL noprint;

     select mean(age), mean(weight) into :AverageAge, AverageWeight from your_data_set;

quit;

Data your_data_set2;

     set your_data_set;

     Age_variance=Age - &AverageAge;

     Weight_variance=weight - &AverageWeight;

run;

[Astounding]

Forgive me if I need another cup of coffee on this one, but ...

If you compute the difference from the mean on each observation, then add up all the differences, doesn't the total have to be zero?

[Reeza]

No because of rounding/floating point error

Yes otherwise :smileysilly: