topic Re: PROC COMPARE Summary Output in SAS Programming

PROC COMPARE Summary Output

MART1 — Thu, 07 Jan 2021 12:16:52 GMT

Hello

I'm started using the (very powerful) PROC COMPARE.

I'd like to generate a summary output and briefsummary does it nicely; however I would like to remove the WARNING and NOTE appearing at the very beginning of the output (in the yellow box on the screenshot).

Is this possible?

Many thanks

Re: PROC COMPARE Summary Output

s_lassen — Thu, 07 Jan 2021 12:26:37 GMT

I think you need to present the code you are using (and possibly some sample data, as datastep code). Otherwise it is not obvious what you are doing.

Re: PROC COMPARE Summary Output

Kurt_Bremser — Thu, 07 Jan 2021 13:04:42 GMT

No. These WARNINGs are there for a reason, alerting you to data that might make your compare incorrect. To get rid of the WARNINGs, either get rid of the duplicate observations, or find an additional ID variable that makes the compares unique.

Re: PROC COMPARE Summary Output

MART1 — Thu, 07 Jan 2021 14:16:48 GMT

Thanks @Kurt_Bremser

all I needed is to know whether there are duplicates, which option briefsummary does. I wouldn't need to know at which observation the duplications occur ; but from what you say this info cannot be excluded from the result.

thanks

Re: PROC COMPARE Summary Output

data_null__ — Thu, 07 Jan 2021 21:47:27 GMT

Perhaps you could use PROC SORT to find dups.

447  proc sort data=sashelp.class out=dups uniout=uni nouniquekey;
448     by sex age;
449     run;

NOTE: There were 19 observations read from the data set SASHELP.CLASS.
NOTE: 4 observations with unique key values were deleted.
NOTE: The data set WORK.DUPS has 15 observations and 5 variables.
NOTE: The data set WORK.UNI has 4 observations and 5 variables.

duplicates

Re: PROC COMPARE Summary Output

MART1 — Fri, 08 Jan 2021 14:48:50 GMT

Hi @data_null__

Thanks for your suggestion,

Actually I would need a few more info than the dups (I know, that's contrary to what I originally said!!).

Basically I'd like to use a few options from PROC COMPARE (i.e. listing all variables where there is a difference), but without the initial WARNINGs and MESSAGEs (it's just cosmetic, but it may confuse the users of the report).

thanks

(I'm trying to create a high level report with )

Re: PROC COMPARE Summary Output

data_null__ — Fri, 08 Jan 2021 15:40:40 GMT

Use PROC PRINTTO to "capture" the default BRIEF report and edit it into the REPORT you need.

@MART1 wrote:

Hi @data_null__

Thanks for your suggestion,

Actually I would need a few more info than the dups (I know, that's contrary to what I originally said!!).

Basically I'd like to use a few options from PROC COMPARE (i.e. listing all variables where there is a difference), but without the initial WARNINGs and MESSAGEs (it's just cosmetic, but it may confuse the users of the report).

thanks

thanks

(I'm trying to create a high level report with )

Re: PROC COMPARE Summary Output

MART1 — Fri, 08 Jan 2021 18:11:08 GMT

Hi @data_null__

I can't find the way to select the outputs from PROC COMPARE to add to PROC PRINT (never used PROC PRINT before unfortunately).

Below is an example with dummy data:

data WORK.TABLE_A;
	input DAY VAL1 VAL2;
	datalines;

20190106 2 452
20190107 5 658
20190108 2 743
20190109 8 44
20190202 9 698
20190202 9 698
;
run;

data WORK.TABLE_B;
	input DAY VAL1 VAL2;
	datalines;

20190106 2 58
20190107 5 658
20190108 2 12
20190109 8 44
20190202 9 698
20190202 9 698
;
run;
title "Test";
PROC COMPARE BASE = TABLE_A COMPARE= TABLE_B 
	briefsummary ;
	ID DAY;
RUN;

(so using PROC PRINT, I'd like to display the 2nd and 3rd boxes, not the one crossed out)

could you please give me an example of how to do this?

many thanks

Re: PROC COMPARE Summary Output

data_null__ — Fri, 08 Jan 2021 21:07:42 GMT

PROC PRINTTO

@MART1 wrote:

Hi @data_null__

I can't find the way to select the outputs from PROC COMPARE to add to PROC PRINT (never used PROC PRINT before unfortunately).

Below is an example with dummy data:
data WORK.TABLE_A;
	input DAY VAL1 VAL2;
	datalines;

20190106 2 452
20190107 5 658
20190108 2 743
20190109 8 44
20190202 9 698
20190202 9 698
;
run;

data WORK.TABLE_B;
	input DAY VAL1 VAL2;
	datalines;

20190106 2 58
20190107 5 658
20190108 2 12
20190109 8 44
20190202 9 698
20190202 9 698
;
run;
title "Test";
PROC COMPARE BASE = TABLE_A COMPARE= TABLE_B 
	briefsummary ;
	ID DAY;
RUN;
(so using PROC PRINT, I'd like to display the 2nd and 3rd boxes, not the one crossed out)

could you please give me an example of how to do this?

many thanks

Re: PROC COMPARE Summary Output

SASKiwi — Fri, 08 Jan 2021 21:51:49 GMT

If you are simply wanting to compare record by record (1 with 1, 2 with 2 etc) between File A and File B, then you can drop the ID statement and that will get rid of your warnings.

Re: PROC COMPARE Summary Output

MART1 — Tue, 12 Jan 2021 10:11:48 GMT

Hi @SASKiwi

I'd like to have the ID in (which is the primary key in my datasets), to see which ones have different variable's values between datasets.

Without ID, it shows the Obs number (example below), but the users of the reports would rather see the IDs.

thanks

Re: PROC COMPARE Summary Output

SASKiwi — Tue, 12 Jan 2021 19:08:14 GMT

@MART1 - Having an ID statement where the ID variable is not unique on both BASE and COMPARE datasets could lead to erroneous comparisons. That is the whole point of the SAS log warnings. An alternative approach might be to add a "counter" variable to your data that might look like this:

Day Counter
1  1
1  2
2  1
2  2
2  3

A combination of both Day and Counter will form a unique key for your data. Then when you do your COMPARE, your ID statement will be: ID Day Counter;