Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: RE: Creating a Non-Self Mean & Proportions


From   "Nick Cox" <[email protected]>
To   <[email protected]>
Subject   RE: st: RE: Creating a Non-Self Mean & Proportions
Date   Tue, 30 Jun 2009 18:57:25 +0100

Thanks for the detailed example, which helps enormously. 

Otherwise I find your reply puzzling. You don't state which of my three guesses are wrong but your example implies to me that all are on the right lines, and so my code is relevant (except that the name -total- is a slip for -fraction- or -proportion-). 

In line with Martin Weiss' suggestion note that 

. tab score stratum 

is a very simple way to get the table you give. 

Nick 
[email protected]

======================
P.S. to save you from some future examiner or reviewer as pedantic as myself I underline that stratum is the singular of strata.  
======================

Emmanuel Koku

Dear, Nick, thanks for attempting to resolve my question, though the
code fragment did not help.  Please, below are 20 cases, covering 3
different strata, as well as their respective values on the index
variable of interest (i.e., score),  for your examination. What I want
is the count or proportion of each value of the index variable (i.e.,
score) in each strata.

Cluster	Hhold#	Line#	Strata	score
46	7	1	1	0
122	20	4	1	0
122	6	2	1	1
46	1	1	1	2
122	11	2	1	1
122	6	1	1	0
122	16	6	2	0
46	13	1	2	1
46	12	1	2	2
122	17	1	2	0
122	8	2	2	1
46	13	3	2	1
122	5	2	2	0
46	6	3	3	1
122	16	2	3	0
122	15	1	3	2
46	4	2	3	0
46	19	3	3	1
122	4	2	3	0
46	20	4	3	2

So from this scratch data we can have the following distribution of
the 3 score values by strata:

ScoreValues        Strata1    Strata2    Strata3   Total				

0                          3              3              3           9
1                          2              3              2           7
2                          1              1              2           4

N                                                                    20

Thus, we can state that across all strata 9 individuals had a score of
0, 7 had a score of 1 and 4 had a score of 2 -- that is we aggregate
or count each of the values of the index variable (Score) across or
over all strata.  In this way I can describe the distribution of the
scores over the strata.

I hope this makes it a bit clearer.  If, not  please let me know and I
can try and provide further clarification.

On Mon, Jun 29, 2009 at 4:30 PM, Nick Cox <[email protected]> wrote:

> A meta-question here is why no one has replied to this since it was
> posted. An answer to that is that trying to work out what is wanted
> requires too much effort. Why not give an example of your data? Why make
> everyone try to imagine what you have?
>
> Anyway, my guesses are:
>
> 1. You have a long data structure with personal id (say -id-) and
> stratum id (say -stratum-). (One stratum, several strata.)
>
> 2. You have a variable (say -score-) with values 0,1,2. (What that
> variable represents, how many negative attitudes the others in the
> stratum have, it seems, is not important here.)
>
> 3. You want to summarize the fractions of 0, 1, 2 in each stratum.
>
> If these guesses are all correct, then one answer is
>
> bysort stratum score : gen freq = _N
> by stratum: gen total = freq / _N
>
> You may also want to -collapse- the data.
>
> If these guesses are wrong, you may need to try again with more
> information.

Emmanuel Koku

> I have created a summed scale of 2 dichotomous items coded originally
> 0 1.  The new variable now has values ranging from 0 to 2.
>
> Since I am using a survey that I want to create summary measures like
> non-self mean or non-self proportions that represent other members of
> the respondents strata but excluding the respondent. Eg. I would want
> to know the proportion of the respondents strata who hold no negative
> attitudes - ie 0 in the summed scale; the proportion who hold only one
> of the negative attitudes, ie, 1 in the summed scale as well as those
> members of the strata who hold both negative values or 2 in the summed
> scale.
>
> Through help received here and from a statafaq article, i have been
> able to create similar summary measures for dichotomous items, but i'm
> stumped with this 3-category variable, and would appreciate any
> assistance.

*
*   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/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index