Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: RE: RE: Fwd: Minimum in 24 hours
From
Paul McCabe <[email protected]>
To
[email protected]
Subject
Re: st: RE: RE: Fwd: Minimum in 24 hours
Date
Fri, 17 Feb 2012 15:48:46 +0000
Nick,
That's perfect - thank you
Best wishes
Paul
On 17 February 2012 15:45, Nick Cox <[email protected]> wrote:
> It's also true that -egen- under -by:- interprets subscript references w.r.t. the entire dataset, and not with reference to groups defined by -by:-. This could be interpreted as a bug or misfeature, but the help does warn you against using explicit subscripting.
>
> Nick
> [email protected]
>
>
> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf Of Nick Cox
> Sent: 17 February 2012 15:24
> To: '[email protected]'
> Subject: st: RE: Fwd: Minimum in 24 hours
>
> Setting aside other details, the Stata function -min(varname)- used in -gen newvar = min(varname)- might be expected just to return values of -varname- as you fed it one argument only. As it happens -min(arg)- is illegal in Stata as two or more arguments are required.
>
> Note that the effect of also specifying any -if- qualifier would to restrict evaluation to observations selected, and not at all to extend comparisons to include other observations, which was your intention. Values for observations not selected by the -if- would just be returned as missing. But, as said, using -min()- with one argument is illegal.
>
> The -egen- function -min()- despite having the same name works quite differently. -egen newname = min(varname)- would put the minimum of -varname- over all observations to a new value (constant) for all observations; conversely you can restrict comparisons to subsets.
>
> Using subscripts with -egen- is dangerous as -egen- can (temporarily) change the -sort- order, and thus subvert your intentions.
>
> What you want can be done in various ways. Here is one. I separate out calculation of the first time for each identifier.
>
> local oneday = 60000 * 60 * 24
> bysort id (time) : gen first = time[1]
> egen minMV = min(MV / (inrange(time, first, first + `oneday')) , by(id)
>
> The trickery used here is easier than may appear at first sight.
>
> 1. It is crucial that -egen-'s -min()- function will accept expressions and is not restricted to working with a varname as argument.
>
> 2. The expression MV / (inrange(time, first, first + `oneday') is evaluated as follows.
>
> inrange(time, first, first + `oneday') will evaluate as 1 if true and 0 if not true.
>
> Then MV / (O or 1) evaluates as missing if the denominator is 0 and MV otherwise.
>
> The effect is to return missings for observations you want ignored, those after the first day, and -min()- does ignore those missings in calculating the minimum.
>
> There is much more discussion in
>
> SJ-11-2 dm0055 . . . . . . . . . . . . . . Speaking Stata: Compared with ...
> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. J. Cox
> Q2/11 SJ 11(2):305--314 (no commands)
> reviews techniques for relating values to values in other
> observations
>
> Nick
> [email protected]
>
> Paul McCabe
>
> I have data relating to admissions to intensive care units with data a
> little like this
>
>
> | id time MV |
> |-------------------------------------|
> 1. | 1128809 26oct2005 04:30:00 11 |
> 2. | 1128809 26oct2005 05:30:00 7 |
> 3. | 1128809 26oct2005 06:30:00 8.3 |
> 4. | 1128809 26oct2005 07:30:00 8.4 |
> 5. | 1128809 26oct2005 08:30:00 8.4 |
> |-------------------------------------|
> 6. | 1128809 26oct2005 09:30:00 8.1 |
> 7. | 1128809 26oct2005 10:30:00 8.8 |
> 8. | 1128809 26oct2005 11:30:00 7.8 |
> 9. | 1128809 26oct2005 17:30:00 7.6 |
> 10. | 1128809 26oct2005 18:30:00 8 |
> |-------------------------------------|
> 11. | 1128809 26oct2005 19:30:00 8 |
> 12. | 1128809 26oct2005 20:30:00 8.1 |
> 13. | 1128809 26oct2005 21:30:00 9.2 |
> 14. | 1128809 26oct2005 22:30:00 8.9 |
> 15. | 1128809 26oct2005 23:30:00 10.4 |
> |-------------------------------------|
> 16. | 1128809 27oct2005 00:30:00 9.6 |
> 17. | 1128809 27oct2005 01:30:00 7.8 |
> 18. | 1128809 27oct2005 02:30:00 7.8 |
> 19. | 1128809 27oct2005 03:30:00 7.7 |
>
> I want the minimum value within the first 24 hours of admission.
> However the function min() returns:
>
> gen minMV=min(MV) if time>=time[1] & time<(time[1]+8.64e7) & id==1128809
> invalid syntax
> r(198);
> And while the egen function works with explicit ids, does not allow
> explicit subscripting when I try to combine with by:
>
> sort id time
> by id:egen minMV=min(MV) if time>=time[1] & time<(time[1]+8.64e7)
> 8469 missing values generated)
> duplicates examples id minMV
> Duplicates in terms of id minMV
> +--------------------------------------------+
> | group: # e.g. obs id minMV |
> |--------------------------------------------|
> | 1 19 1 1128809 7 |
> | 2 593 20 1128809 . |
> | 3 22 633 1128817 10 |
> | 4 437 613 1128817 . |
> | 5 18 1072 1128827 . |
> |--------------------------------------------|
> | 6 474 1090 1128828 . |
> | 7 73 1564 1275019 . |
> | 8 118 1637 1613911 . |
> | 9 282 1755 1621603 . |
> | 10 246 2037 1630155 . |
> |--------------------------------------------|
> | 11 465 2283 1633930 . |
> | 12 356 2748 1660876 . |
> | 13 64 3104 1667289 . |
> | 14 52 3168 1687505 . |
> | 15 754 3220 2109887 . |
> |--------------------------------------------|
> | 16 46 3974 2124610 . |
> | 17 47 4020 2129680 . |
> | 18 56 4067 2141764 . |
> | 19 149 4123 2147829 . |
> | 20 402 4272 2150550 . |
> |--------------------------------------------|
> | 21 676 4674 2189948 . |
> | 22 304 5350 2201604 . |
> | 23 955 5654 2202648 . |
> | 24 32 6609 2208416 . |
> | 25 641 6641 2212494 . |
> |--------------------------------------------|
> | 26 1229 7282 2217835 . |
> +--------------------------------------------+
> (ID 1128817 was admitted at the same time as ID1128809 which is why
> minMV is created for them)
>
> Can anyone give advice on how to combine a function, by, and the time selection?
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/