About Yegen

Yegen · ‎06-07-2017

@Jagadishkatam, when I run your code on the following sample, the output does not look correct. In particular, the observation with the ID 17311711 should have not been selected. Do you know where the mistake is happening? I would also like to know how to write this code using PROC SQL. Thanks. DATA work.sample; format group industry id control; INPUT ID Industry Group $ Control; DATALINES; 55203610 2542 AB 1 16124130 8062 AB 1 16124130 8062 AB 0 80479512 2062 AB 1 70321918 2062 AB 1 17312410 6712 AB 1 17311710 6712 AB 0 17311711 2082 AB 0 74623123 2082 AC 1 24245321 2082 AC 0 ; RUN; proc sql; create table want as select id,group from sample where industry in (select distinct industry from sample group by industry having count(Industry )>1) group by industry,id having count( Industry)<=1 order by group; quit;

Yegen · ‎06-07-2017

Thanks @Shmuel, this is a great way of identifying the observations! Although the code does exactly what I was looking for, I am not quiet sure why you had to create a new column of lagged variables that are not within the same group or industry. Could you please elaborate on that line?

Yegen · ‎06-07-2017

If you only want to select random As, Bs, and Cs, the following code will take care of it: data work.want; do i=1 to 10; random_ABC = substr('ABC',ceil(ranuni(0)*3),1); output; end; run;

Yegen · ‎06-07-2017

@ChrisNZ thanks for this helpful suggestion!

Yegen · ‎06-07-2017

Other than just using a left join with the original dataset, is there another easy way to identify such observations? I am trying to learn any alternative methods, so your suggestion will be very helpful. What I was thinking of is just doing a left join (where on the left side the original data is and the right side the output of the code) and then identifyin the ones that are having missing values after the merge.

Yegen · ‎06-07-2017

I am trying to move completely to SAS (from other statistical softwares) but I am having a difficult time getting used to "row by row" operations. For example, in this particular case, I am trying to compare rows within a given group and keep the rows that have not the identical ID number but are within the same industry and group. Here is my sample: DATA work.sample; INPUT ID Industry Group $ Control; DATALINES; 55203610 2542 AB 1 16124130 8062 AB 1 16124130 8062 AB 0 80479512 2062 AB 1 70321918 2062 AB 1 17312410 6712 AB 1 17311710 6712 AB 0 74623123 2082 AC 1 24245321 2082 AC 0 ; RUN; Here is what the output needs to look like: Data work.want; INPUT ID Group $; DATALINES; 80479512 AB 70321918 AB 17311710 AB 17312410 AB 74623123 AC 24245321 AC ; RUN; Here is my code: proc sort data=work.sample; by Group Industry ID; quit; DATA work.identified; set work.sample; by Group; if lag(ID)^=ID and lag(Industry)=Industry then keep=1; else keep=0; RUN; DATA work.identified; set work.identified; if keep=1; Run; With that code, I am incorrectly identifying some of the first observations within a given group and the lagged observations that have a subsequent non-identical ID which is within the same group and industry. For instance, with the code above, I have misidentified 70321918, 17311710, and 24245321. Any suggestions / help will be greatly appreciated.

Yegen · ‎06-05-2017

I see. That is good to know. So there will be only one observation (i.e., the last announcement) whenever fdate is not between announcements. Thanks.

Yegen · ‎06-05-2017

Thanks, @Ksharp. Your help is much appreciated!

Yegen · ‎06-05-2017

@Xiaoningdemao, glad to see that @SuzanneDorinskiwas able to solve the issue. Cheers

Yegen · ‎06-02-2017

Hi @Xiaoningdemao, Great that you have moved to SAS Studio, it's a very convenient way of using SAS (especially if one is a Mac user or wants to be more cloudy)! I can refer you to @Cynthia_sas's answer to a related question: https://communities.sas.com/t5/SAS-Analytics-U/How-do-I-turn-off-output-in-Studio-University-Edition/td-p/277820 Although the person who asked the question was wondering how to turn off the output / result, you can use ODS to change the path of your output and save it directly in a folder. That way, you can access your output right away even if it is not saved in your Results window. Additionally, here is the link that provides some information on ODS and how to change the output environment: http://support.sas.com/documentation/cdl/en/webeditorug/68828/HTML/default/viewer.htm#n0lyxx45gr31pan1vmm8z5f952l3.htm For instance, you can use the following code if you want to save your result in a given folder (e.g., in your Output folder). ods html path='SPECIFY LOCATION OF YOUR FOLDER'; I hope this helps a bit, but I think the links provided above should give you some guidance on how to save your output in SAS Studio and get around of losing your results. Cheers

Yegen · ‎06-01-2017

Perfect, thanks for pointing that out, @Reeza.

Yegen · ‎06-01-2017

I know the following code works well, but does anyone have alternative suggestions to write a more efficient code that replaces zeros with missing values? Here is a sample dataset: data work.sample; infile cards expandtabs truncover; input ID shrout shrout1 shrout2; cards; 26232409 0 2 50997000 26232409 0 3 50997000 26232409 53598 0 50997000 10623630 10097 2 0 5856168 . . . ; run; Here is the code: data work.sample; modify work.sample; array vars{*} shrout shrout1 shrout2; do i = 1 to dim(vars); if vars{i}=0 then call missing(vars{i}); end; run;

Yegen · ‎06-01-2017

Thanks for the clarification, @ballardw. Also, thank you for the helpful link, @Reeza.

Yegen · ‎06-01-2017

Sorry this is the correct output table (I have updated the table in the question as well): shares shrout shrout1 shrout2 WANT 26232409 53598 50997 50997000 0.000489428878 10623630 10097 . 16140743 0.001052157076 12985791 . . 67659160 0.00019193 3653140 . 21212 21212948 0.00017222 5856168 . 41221 . 0.000142068 5856168 . . . . @kiranv_ what do you think about the modified version of your code? If it works, maybe you can update your reply and I can accept it as an answer so that others can make use of it as well. proc sql; create table work.want as select shares, case when coalesce(shrout,shrout1) ^=. then shares/(coalesce(shrout,shrout1)*1000*1000) else shares/(shrout2*1000) end as share_rates from work.sample; quit; This produces the exact values. Could you please provide some details about what the "coalesce" statement does? Does it take into account for the missing value order? So, let's say if both shrout and shrou1 are not missing, then will the code use shrout instead of shrout1? In other words, what exactly does the "coalesce" statement do?

Yegen · ‎06-01-2017

@kiranv_ thanks for your reply. Your code does not produce the correct output. I see what you are trying to do, but the formulas used are incorrectly specified. Here is a summary of which method needs to be used when: If Then Shrout ^= . shares / (shrout*1000*1000) Shrout = . And Shrout1^=. shares / (shrout1*1000*1000) Shrout = . And Shrout1=. shares / (shrout2*1000)

Online Status	Offline
Date Last Visited	‎06-01-2020 06:53 PM

Re: Rank Transformation with mean=0 and std=1

Re: Rank Transformation with mean=0 and std=1

Re: Rank Transformation with mean=0 and std=1

Rank Transformation with mean=0 and std=1

Re: Error when using &md

Error when using &md

Re: Is it possible to absorb fixed effects when using Proc Surveyreg?

Re: Compress empty space of all variables

Re: Compress empty space of all variables

Re: Compress empty space of all variables

Re: Rank Transformation with mean=0 and std=1

Re: Rank Transformation with mean=0 and std=1

Re: Compress empty space of all variables

Re: Compress empty space of all variables

Re: Is it possible to absorb fixed effects when using Proc Surveyreg?

Re: Is it possible to absorb fixed effects when using Proc Surveyreg?

Re: Is it possible to absorb fixed effects when using Proc Surveyreg?

Re: Compress empty space of all variables

Re: Assigning group identifiers after merger

Assigning group identifiers after merger

Re: Row by row comparison within group

Re: Row by row comparison within group

Re: set variable equal to a randomly chosen value from a list

Re: Is there an alternative to in( ) in Proc SQL?

Re: Is there an alternative to in( ) in Proc SQL?

Row by row comparison within group

Re: Is there an alternative to in( ) in Proc SQL?

Re: How to keep variables that are not in the SELECT DISTINCT statemen...

Re: How to access old output result

Re: How to access old output result

Re: Replacing zeros with missing values

Replacing zeros with missing values

Re: Multiple if then / else then statement in one argument

Re: Multiple if then / else then statement in one argument

Re: Multiple if then / else then statement in one argument