Dear Statalisters,
A few days back, I posted a question about restructuring multilevel
data. Nick Cox suggested I send the data structure rather than the
narrative explanation I originally posted. For some reason, when I
sent the message below, the original message was posted instead. My
apologies for the repetition, but here is the response to Nick's
request for clarification. The original message & response appears
below. Again, many thanks in advance for your suggestions.
Steve Harvey
----------------------------------------------------------
My data is structured as follows. I'm putting the varnames in a column
here rather than a row because of the number of vars. involved.
prov#_id is each individual provider's id number.
prov#_lab, prov#_del, prov#_m, prov#_nn are all dummy variables
indicating, respectively, whether each provider participated in labor
monitoring, delivery, post partum care for the mother, and post partum
care for the neonate.
obs_no (numeric 1-n)
[demographic var list]
prov1_id (numeric)
prov1_lab (0, 1)
prov1_del (0, 1)
prov1_m (0, 1)
prov1_nn (0, 1)
prov2_id (numeric)
prov2_lab (0, 1)
prov2_del (0, 1)
prov2_m (0, 1)
prov2_nn (0, 1)
.
.
.
prov8_id (numeric)
prov8_lab (0, 1)
prov8_del (0, 1)
prov8_pm (0, 1)
prov8_pnn (0, 1)
[explanatory varlist]
I've used reshape in other situations. I don't think it will work here
because I need to include the demographic and explanatory variables in
each observation, but generate multiple observations per case (woman
delivering) so that each observation contains information about only
one provider.
I hope this explanation is more clear. Thanks again for your
suggestions.
Steve
---------------------------------
Date: Wed, 15 Oct 2003 18:04:58 +0100
From: "Nick Cox" <[email protected]>
Subject: st: RE: Restructuring & merging multi-level data
Steven A. Harvey, MHS, PhD(c)
> I am attempting a multi-level analysis of data on the
> performance of
> skilled birth attendants at managing labor and delivery in
> 4 countries. My
> data set on performance is structured with the case (i.e.,
> the woman
> delivering) as the unit of analysis. For each case, I have up to 8
> variables to record the ID numbers of providers who
> participated (i.e., up
> to 8 providers might have participated in some phase of any
> given case). I
> also have 4 dummy variables indicating in which of 4 stages
> of labor &
> delivery each provider participated.
>
> My explanatory variables are collected at the provider
> level. In these
> datasets, I have an individual provider ID # plus
> information about the
> provider's experience, scores on tests, perspectives on the
> quality of the
> health facility in which he/she works, etc.
>
> I need some mechanism in Stata that will help me reorganize
> the first
> dataset so that I have one record for each time a provider
> participated in
> a case (provider x case). Then I want to match this with
> the explanatory
> variables based on provider ID#, presumably using Stata's
> merge command.
>
> Other than reorganizing the data by hand, I am at a
> complete loss about how
> to approach this. I would greatly appreciate any suggestions.
>
It is usually much easier to understand a problem
from an example of data than from a verbal description.
My reading of this, however, is that you need a -reshape-
to long, followed as you say by a -merge-. You may well
need some renaming of variables before you can -reshape-.
Check out [R] reshape and also various extras at
http://www.stata.com/support/faqs/data/reshape3.html
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/
