Nick,
Following your suggestion, I have -reshape- the data. This is how it
looks now:
id chilno dfsize nboy alive b boyno cno
1 1 4 2 4 1 1 1
1 2 4 2 4 2 1 2
1 3 4 2 4 1 2 3
1 4 4 2 4 2 2 4
1 5 4 2 4 . 2 4
1 6 4 2 4 . 2 4
1 7 4 2 4 . 2 4
2 1 4 2 3 1 1 1
2 2 4 2 3 . 1 1
2 3 4 2 3 2 1 2
2 4 4 2 3 . 1 2
2 5 4 2 3 1 2 3
2 6 4 2 3 . 2 3
2 7 4 2 3 . 2 3
dfsize: desired family size
nboy : number of boys alive in the family
alive : noumber of children alive
b : 1(boy), 2(girl), .(dead or not born)
boyno : first boy or second boy, etc.
cno : first child or second child, etc. (only alive children)
1. I want to first generate the following variables:
local i 1
while `i' <= 7 { \\There are a maximum of 7 children in any family
local j 1
while `j'<= `j' {
gen p`i'`j'=0
local j= `j' + 1
}
local i = `i' + 1
}
local k = `nboy'
local i 1
forvalues i=1(1)`k' {
replace p`dfsize'`k' = (0.5)^`alive'
}
I can do this step.
2. Now I want to replace the value of p`dfsize'`k' with 0 whenever the
kth (k runs from 1 to `nboy') boy is not the Nth child. For instance,
for the family with id==1, I would want: p41=0, p42=0, p43=(0.5)^4. For
the second family (id==2), I would want: p51=0, p52=(0.5)^3, p53=0,
p54=0. (Note that I want the p* variables to have a single value for
each family.)
This is where I am stuck.
2. Later, I want to get back to the -wide- format because I want to use
the p* variables for my ML estimation.
Any suggestions will be appreciated.
Deepankar
On Mon, 2006-06-05 at 14:59 +0100, n j cox wrote:
> It is indeed a good idea to tell us the actual problem.
>
> You would be well advised to -reshape- your data from wide
> to long. The other way lies mostly analytic awkwardness and
> programming pain. My guess is that your guess about
> needing nested loops is incorrect.
>
> Assume an identifier -id- as well as the variables
> you report.
>
> keep id dfsize b*
> reshape long b , i(id) j(childno)
>
> Then you get which boy it is by
>
> bysort id (childno) : gen boyno = sum(b == 1)
>
> Then you can relate -boyno- and -childno-.
>
> You may then need to -reshape- back.
>
> Nick
> n.j.cox@durham.ac.uk
>
> DEEPANKAR BASU
>
> Thanks for your comments. My actual programming problem is much more
> complicated than just generating p`x'`y' by checking whether `x' == `y'.
> I had given that example as a prototype of the problem I am facing with
> the "if" statement. I could re-phrase my question as: can I do what
> Keith has suggested by using the "if ... else" statement? I give details
> of my actual problem below.
>
> I am working with a (survey-based) dataset which has the following
> variables (related to fertility behaviour) among others:
>
> 1. dfsize: desired family size
> 2. b1: sex of the first child (1=boy, 2=girl, and "missing" if no first
> child in the family)
> 3. b2: sex of second child
> ...
> 11. b9: sex of ninth child
> 12. nboy: number of boys alive in the family
>
> I want to do the following: for each family (observation), I want to see
> whether the k_th boy is the the N_th child, where k goes from 1 to (N-1)
> and N is "dfsize" (the desired family size). If it is the case that the
> k_th boy is the N-th child, then I want to have:
> p`N'`k'=(0.5)^`N';
> Otherwise, I want to have:
> p`N'`k'=0
>
> Note that I will have several nested loops within an outer "if...else"
> statement. The question is: can I do the whole thing differently,
> without using the "if...else" statement?
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/