Date:         Wed, 21 Jul 2010 13:44:27 -0400
Reply-To:     Ya Huang <ya.huang@AMYLIN.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Ya Huang <ya.huang@AMYLIN.COM>
Subject:      Re: summary statistics of all variables
Comments: To: Mike Zdeb <msz03@ALBANY.EDU>

I'e been thinking similarly. But instead of parsing the text file,
why don't out a file that can be read by SAS directly? So, I tried
ods xml:

options nodate nonumber;
ods xml file="c:\temp\junk.xml" style=minimal;

proc means data=sashelp.class nmiss n mean median max min;
var _numeric_;
title;
run;

ods xml close;

libname stat xml "c:\temp\junk.xml";

data stat;
 set xx.junk
run;

Unfortunately, the xml file is a bit too complicate to be read
by xml engine directly without using XMLMAP thingy.

I also tried tagsets.excelxp, hoping to get the excel file so I can
read in with proc import or excel libname engine. BUT, to my surprise,
the layout is different! It's the output dataset layout.

Ya


On Wed, 21 Jul 2010 13:25:10 -0400, Mike Zdeb <msz03@ALBANY.EDU> wrote:

>hi ... given the observation on standard MEANS output, rather than ODS
OUTPUT
>how about output to a file then read the results into a data set ...
>
>* make some missing data in sashelp.class;
>data class;
>set sashelp.class;
>if ranuni(999) le .3  then call missing(of _all_);
>run;
>
>* redirect output to file;
>filename means temp;
>proc printto print=means;
>run;
>
>proc means data=class n nmiss min max;
>run;
>
>proc printto;
>run;
>
>* read proc means output;
>data stats;
>infile means truncover;
>input variable : $32. (n nmiss min max) (??);
>if max;
>run;
>
>proc print data=stats;
>run;
>
>Obs    variable     n    nmiss     min    max
> 1      Age        13      6      11.0     15
> 2      Height     13      6      51.3     69
> 3      Weight     13      6      50.5    133
>
>--
>Mike Zdeb
>U@Albany School of Public Health
>One University Place (Room 119)
>Rensselaer, New York 12144-3456
>P/518-402-6479 F/630-604-1475
>
>> From the replies you have received it appears that a data set with the
>> structure you require is somewhat difficult to achieve.  What with all
>> that transposing and the rest.
>>
>> Interestingly the output closely resembles the printed output from PROC
MEANS.
>>
>> The MEANS Procedure
>>
>>                 N
>> Variable     Miss      N         Minimum         Maximum
>> Age           395     19      11.0000000      16.0000000
>> Height        395     19      51.3000000      72.0000000
>> Weight        395     19      50.5000000     150.0000000
>> Stores         19    395       1.0000000      41.0000000
>> Sales          19    395     325.0000000      1298717.00
>> Inventory      19    395     374.0000000      2881005.00
>> Returns        19    395      10.0000000        57362.00
>>
>> But when you ask for this data via ODS OUTPUT the data structure is
>> very different. And the transposing begins.
>>
>> The ODS output from PROC CORR does resemble the desired layout but you
>> don't get to specify statistics and NMISS is not available, but could
>> be calculated.
>>
>> ods listing close;
>> proc corr nocorr data=test;
>>    ods output "Simple Statistics"=stats /*"Variables
Information"=Varinfo*/;
>>    run;
>> ods listing;
>> Ods trace off;
>> proc print data=stats;
>>    run;
>>
>> Obs    Variable         NObs        Mean      StdDev           Sum
>>     Min         Max
>>
>> 1     Age                19        13.3        1.49     253.00000
>>  11.00       16.00
>> 2     Height             19        62.3        5.13          1184
>>  51.30       72.00
>> 3     Weight             19       100.0       22.77          1901
>>  50.50      150.00
>> 4     Stores            395        11.6        8.87          4601
>>   1.00       41.00
>> 5     Sales             395     85700.2    129107.2      33851566
>> 325.00     1298717
>> 6     Inventory         395    250898.9    351514.6      99105051
>> 374.00     2881005
>> 7     Returns           395      2967.3     4611.74       1172092
>>  10.00    57362.00
>>
>>
>> Since the statistics you request are included in the DEFAULT output
>> data set from PROC SUMMARY and NMISS can be easily calculated this is
>> another possibility.
>>
>> proc summary data=test nway;
>>    var _numeric_;
>>    output out=stats;
>>    run;
>> proc transpose out=test2;
>>    by _type_ _freq_;
>>    id _stat_;
>>    run;
>> data test2;
>>    set test2;
>>    nMiss = _freq_ - N;
>>    run;
>> proc print;
>>    run;
>>
>>
>>
>>
>> On 7/20/10, Jeff <zhujp98@gmail.com> wrote:
>>> I have a datasets with about 200 numeric variables.
>>> I want to create a dataset containing Nmiss, N, Max and Min for all
>>> variables.
>>> Is there a good ways to do that?
>>> The resulting dataset should looks like:
>>>
>>> varname #ofmissing   N          max min
>>> qqq         543          123472    80    0
>>> sdd
>>> dsdsd
>>> ssdsdd
>>>
>>> Thanks.
>>> Jeff
>>>
>>