RE: Calculation person-year.

[art297]

Phan,

I failed to mention that the code I suggested requires SAS 9.3.  I presume you are on an older version.

[PhanS]

Dear Art,

Thanks for the follow-up. Indeed, I am working on 9.3.

In fact, this message from log file: character values have been converted  to numeric values at the places given by ... referred for another (I'm sorry to confuse you).

Lately, I will send you a log file that may help you to understand the issue when I ran the code.

I am in Vancouver, a bit inconvenience at my workplace for Internet.

Phan

[art297]

Not a problem.  I'm in Toronto thus at least your response won't have to go through Customs.

[PhanS]

Dear Art,

Please kindly see the log. I will see a stat person tomorrow morning and will fu you.

8617  /*person-year*/

8618  py_dead1=yrdif(farvdt, death_date, 'act/act');

8619  py_dead2=datdif(farvdt, death_date, 'act/act');

8620

8621

8626  run;

NOTE: Invalid argument to function YRDIF at line 8618 column 10.

NOTE: Invalid argument to function DATDIF at line 8619 column 10.

Thanks,

Phan

[art297]

Are both farvdt and death_date SAS date fields?  All three of the following work for me:

data have;

  informat farvdt death_date date9.;

  input farvdt death_date;

  /*person-year*/

  py_dead1=yrdif(farvdt, death_date, 'act/act');

  py_dead2=datdif(farvdt, death_date, 'act/act');

  py_dead3=yrdif(farvdt, death_date, 'age');

  cards;

1jan2010 10may2012

10jun2011 25oct2012

;

[PhanS]

Dear Art, good morning!

Pls. kindly see the date SAS. I will fu with you late today, I hope to show the issue to my colleague. Thanks.

Phan

                                                   DEATH_

                       Obs     ID FARVDT              DATE

                        16     310 01MAR2010            .

                        17     340 28APR2008            .

                        18     350 22NOV2006    27MAR2008

                        19     360 23FEB2004            .

                        20     370 07OCT2009            .

[art297]

You would get the error due to the missing dates.  What date would you base person_years on for someone who hasn't died?

[PhanS]

Dear Art,

You are absolutely correct!

I use another date (as this dataset is a bit confusion) the dead_date do not reflect for death and I am asked to use fu_time. And the code work fine.

I much appreciate and thank you very much, Art, and thanks everyone's times for discussion.

Phan S.