Bookmark and Share

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: forvalues foreach or while loop for sequential variables


From   Robbie Katz <robbiekatz05@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: forvalues foreach or while loop for sequential variables
Date   Thu, 31 Oct 2013 15:23:26 -0400

Thanks Roberto, A Loumiotis, and Nick!

I should have mentioned this earlier but, the variables are date
variables, and the data set was previously in a long format. The
problem 1, identified by Nick will not occur because I sorted the
variable hosp, before reshaping to wide, so something like "6,9,6"
will not occur. However, something like "5,5,5," could occur, and the
code that I had would miss that, but looks like Nick's solution is apt
to avert this problem.

Nick I think I sort of answered your question above, but I need to
throw away the extra dates if they are repeated as they don't add any
extra information.

Thank you guys a bunch!



On Thu, Oct 31, 2013 at 5:12 AM, Nick Cox <njcoxstata@gmail.com> wrote:
> (Revised version of previous)
>
> This picks up the syntax Robbie was missing, namely
>
> `=exp'
>
> for some expression exp, but it won't work entirely as desired.
> Suppose -hospital7- -hospital8- -hospital9- are all 42. Then
> -hospital8- is changed to missing because it is the same as
> -hospital7-, but next time around the loop -hospital9- of 42 is
> compared with missing and nothing is changed.
>
> Roberto pointed out this problem too.
>
> The following should be an improvement for a solution without -reshape-.
>
> clear
>  input obs hosp14  hosp15 hosp16
>  1   1      5    5
>  2   6      9    6
>  3   8      1    3
>  4   5      5    5
>  5   2      2    2
>  6   2      9    1
>  7   1      1    1
>  8   3      1    7
>  9   3      0    4
>  end
>
> gen previous = hosp14
> gen temp = .
>
> qui forval j = 15/16 {
>        replace temp = hosp`j'
>        replace hosp`j' = . if hosp`j' == previous
>        replace previous = temp
> }
>
> drop previous temp
>
> That said,
>
> 1. The non-missing values left will not be guaranteed unique or
> distinct. Hospitals 6, 9, 6 visited in that order will be left as is
> and 6 will occur twice.
>
> 2. Throwing away information is always risky. Why do you want to do this?
>
> 3. As these are in essence panel data, Roberto's implication that in
> many ways you are better off with a long data structure (I don't like
> the word "format" here) is worth pondering.
>
>> Nick
>> njcoxstata@gmail.com
>>
>>
>> On 31 October 2013 07:32, A Loumiotis <antonis.loumiotis@gmail.com> wrote:
>>> another way without defining a new local would be:
>>>
>>> replace hosp`num'=. if hosp`num'==hosp`=`num'-1'
>>>
>>> Antonis
>>>
>>> On Thu, Oct 31, 2013 at 8:52 AM, Roberto Ferrer <refp16@gmail.com> wrote:
>>>> Elaborating a bit more, according to your example, you intend
>>>> something along the lines of:
>>>>
>>>> *-------- begin code ---------
>>>> clear
>>>> set more off
>>>>
>>>> input obs hosp14  hosp15 hosp16
>>>> 1   1      5    5
>>>> 2   6      9    6
>>>> 3   8      1    3
>>>> 4   5      5    5
>>>> 5   2      2    2
>>>> 6   2      9    1
>>>> 7   1      1    1
>>>> 8   3      1    7
>>>> 9   3      0    4
>>>> end
>>>>
>>>> list
>>>>
>>>> * You have this in a loop
>>>> replace hosp15 = . if hosp15 == hosp14
>>>> list
>>>> *--------- end code ----------
>>>>
>>>> but this compares only contiguos hospitals (14 with 15 and 15 with 16) but not
>>>> hospitals that are "further apart" (14 with 16). This is the case with
>>>> observation 2 in the example data I created. Your loop misses this comparison,
>>>> which I assume is important, i.e., you want hosp16 to equal .
>>>> for that observation because hosp14 already takes the value of 6. If
>>>> the assumption
>>>> is true, this is very problematic.
>>>>
>>>> That said, now take a look at what your variables are evaluating to
>>>> inside the loop
>>>> (use -display- for this):
>>>>
>>>> *-------- begin code ---------
>>>> clear
>>>> set more off
>>>>
>>>> input obs hosp14  hosp15 hosp16
>>>> 1   1      5    5
>>>> 2   6      9    6
>>>> 3   8      1    3
>>>> 4   5      5    5
>>>> 5   2      2    2
>>>> 6   2      9    1
>>>> 7   1      1    1
>>>> 8   3      1    7
>>>> 9   3      0    4
>>>> end
>>>>
>>>> forvalues num=15/16 {
>>>>           display "hosp`num'" // your LHS
>>>>           display "hosp``num'-1'" // your RHS
>>>>           replace hosp`num'=. if hosp`num'==hosp``num'-1'
>>>> }
>>>>
>>>> *--------- end code ----------
>>>>
>>>> The LHS is OK, but the RHS is always evaluating to "hosp". You do not
>>>> specify your
>>>> error in your post (you mention "it didn't work" and you should be
>>>> specific about it)
>>>> but here I get this:
>>>>
>>>> hosp ambiguous abbreviation
>>>> r(111);
>>>>
>>>> Stata doesn't know if you're refering to hosp14, hosp15 or hosp16, and
>>>> so it halts
>>>> with an error.
>>>>
>>>> Because I mess up easily with all the quotes (maybe someone will
>>>> clarify this point),
>>>> my strategy is to declare a new local and then use it for the RHS:
>>>>
>>>> *---
>>>> forvalues num=15/16 {
>>>>           local newnum = `num' - 1
>>>>           replace hosp`num' = . if hosp`num' == hosp`newnum'
>>>> }
>>>> *---
>>>>
>>>> I hope this helps.
>>>>
>>>> On Thu, Oct 31, 2013 at 1:42 AM, Roberto Ferrer <refp16@gmail.com> wrote:
>>>>> Robbie,
>>>>>
>>>>> Here is one way avoiding the loop and using -reshape-.
>>>>>
>>>>> *-------begin code ---------
>>>>>
>>>>> clear
>>>>> set more off
>>>>>
>>>>> input obs hosp14  hosp15 hosp16
>>>>> 1   1      5    5
>>>>> 2   6      9    6
>>>>> 3   8      1    3
>>>>> 4   5      5    5
>>>>> 5   2      2    2
>>>>> 6   2      9    1
>>>>> 7   1      1    1
>>>>> 8   3      1    7
>>>>> 9   3      0    4
>>>>> end
>>>>>
>>>>> list
>>>>>
>>>>> * Reshape to manipulate
>>>>> reshape long hosp, i(obs) j(hnum)
>>>>>
>>>>> sort obs hosp hnum // important
>>>>> gen hosp2 = hosp
>>>>> by obs: replace hosp2 = . if hosp[_n] == hosp[_n-1]
>>>>>
>>>>> drop hosp
>>>>> rename hosp2 hosp
>>>>>
>>>>> * Take back to original form
>>>>> reshape wide hosp, i(obs) j(hnum)
>>>>> list // compare this with original input
>>>>>
>>>>> *------- end code ---------
>>>>>
>>>>> On Thu, Oct 31, 2013 at 12:41 AM, Robbie Katz <robbiekatz05@gmail.com> wrote:
>>>>>> I have a wide format data with variables with same text and different
>>>>>> numbers attached to them. The variables look like this: hospital1
>>>>>> hospital2 hospital3...hospital15. I want only unique values across
>>>>>> these variable for any one observation. For instance if both hospital
>>>>>> 15 and hospital14 have the same value say, 10, then I want to delete
>>>>>> 10 and replace it with "." for hospital15. Please let me know how I
>>>>>> can run a loop to achieve this.
>>>>>>
>>>>>> I tried the following but it didn't work:
>>>>>> forvalues num=2/15 {
>>>>>>           replace hospital`num'=. if hospital`num'==hospital``num'-1'
>>>>>>   }
>>>>>>
>>>>>> Please help,
>>>>>> Thanks!
>>>>>> *
>>>>>> *   For searches and help try:
>>>>>> *   http://www.stata.com/help.cgi?search
>>>>>> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index