Jun Xu posted twice, here labelled (1) and (2):
(1)
> puzzled by a problem when writing an ado file. Suppose I have
> var1 var2 var3 var4 var5......vark, and I want to do the
> following loop:
> ************************************************************
> *********
> var1
> var2
> var3
> ...
> ...
> ...
> vark
> var1 var2
> var1 var3
> var1 var4
> ...
> ...
> var1 vark
> var2 var3
> var2 var4
> ...
> ...
> var(k-1) vark
> ...
> ...
> var1 var2 var3
> var1 var2 var4
> var1 var2 var5
> ...
> var1 var2 var3....vark
> ******************************************************************
> Basically, what I want to do is like step wise exhausting
> all combinations
> in a systematic way from univariate, bivariate, trivariate, to
> multivariate....Or, I can say for every variable in the
> variable list, there
> is indicator variable associated with it. I either take
> this variable in or
> out in each run. And there should be 2^k possibilities. I
> have no idea how
> to handle that. COuld anyone give me some hint? Many
> thanks in advance.
(2)
I think I might not have explained my problems clearly. I have k
indicator
variables (coded as 1 or 0) and I would like to know the response
patterns
(for example for latent class analsis) to these k variables. For
example,
var1 var2 var3 ....vark
1 0 0 0
0 1 0 0
...
1 1 0 0
...
...
...
1 1 1 1
I would like to know for each response pattern, how many cases are
there,
and programmed into an ado file. My key problem here is how to run
through
all the combinations (univariate, bivariate, and trivariate)
One posibility is that I used the following cods (or reviced version
to fit
into an ado file)
******************************************************************
clear
for num 1/6: set obs 100\ gen xX=invnorm(uniform()) \ gen DxX=xX>0.6
gen pattern=0
local i=1
while `i'<6 {
replace pattern=pattern+Dx`i'*10^(6-`i')
local i=`i'+1
}
aorder
list Dx1-Dx6 pattern
sort pattern
list pattern
gen count=1
collapse (sum) count, by(pattern)
***********************************************************
The resulting data matrix looks like:
============================
pattern count
0 16
10 10
100 5
110 6
1000 8
1010 7
1100 2
1110 1
10000 11
10010 3
10100 2
10110 2
11000 2
11010 2
11100 1
100000 7
100010 1
100100 2
101000 4
101010 1
110000 1
110010 2
110100 2
111000 1
111010 1
=================================
Here the problem is that it only presents the response pattern that
has at
least one case and it's hard to handle its order (now is list in
numerical
order: from small to big)
But what if I want to go through "each" combination (2^k possible
ways) in a
sysmatic way and list all response pattern freqeuncy though some of
them
have zero cases. What I meant by a systematic way is like:
*********************************************************************
var1
var2
var3
...
...
...
vark
var1 var2
var1 var3
var1 var4
...
...
var1 vark
var2 var3
var2 var4
...
...
var(k-1) vark
...
...
var1 var2 var3
var1 var2 var4
var1 var2 var5
...
var1 var2 var3....vark
******************************************************************
or in binary coding
****************************************************************
1 0 0 0 0 .....0
0 1 0 0 0 .....0
0 0 1 0 0 .....0
...
...
...
0 0 0 0 0 .....1
1 1 0 0 0 .....0
1 0 1 0 0 .....0
1 0 0 1 0 .....0
...
...
1 1 1 0 0 .....0
1 0 1 1 0 .....0
...
...
...
...
1 1 1 1 1 .....1
***********************************************
Here I didn't present some summarize command that could grab case
number for
that response pattern. But basically I will run through each
combination
and calculate the frequency for that particular combination though
there
might be zero cases. Thanks a lot
>>> To get a tabulation of patterns with some instances,
egen all = concat(var1-vark)
tab all
The following program suggests some possible lines of attack.
program permlist, rclass
version 8
syntax varlist
tokenize `varlist'
local nvars : word count `varlist'
local imax = 2^`nvars' - 1
forval i = 1 / `imax' {
qui inbase 2 `i'
local which : di %0`nvars'.0f `r(base)'
local vars
forval j = 1 / `nvars' {
local char = substr("`which'",`j',1)
if `char' {
local vars "`vars'``j'' "
}
}
local vlist `"`vlist'"`vars'" "'
}
local varlist
forval i = 1 / `nvars' {
foreach w of local vlist {
local nv : word count `w'
if `i' == `nv' {
local varlist `"`varlist'"`w'" "'
}
}
}
return local varlist `"`varlist'"'
end
I use the undocumented -inbase- command
to get the binary equivalent of 1 ... 2^k - 1 (I omit the null case
in which none of the variables are chosen).
Nick
[email protected]
*
* 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/
