st: Re: st: Constructing an Asset Index in Stata‹Discrepancy between predict command and hand calculations

[Friedrich Huebler <fhuebler@gmail.com>]

Jennifer,

Documentation for -predict- after -factor- is under "factor
postestimation" in the Multivariate Statistics Reference Manual.

Does it matter that the minimum and maximum scores are different? You
only need the asset index to rank households so that they can be
assigned to wealth quintiles. The absolute values are irrelevant, as
long as the relative ranking of all households is the same.

Friedrich

On Mon, Jan 3, 2011 at 1:24 PM, Jennnifer Callaghan <jcallagh@jhsph.edu> wrote:
> Hello and Happy New Year,
>
> I am trying to find out what calculations Stata uses to calculate an asset
> index using the predict command.  I am following the commands given by the
> World Bank ( in the book Analyzing Health Equity Using Household Survey
> Data, by O'Donnell, Van Doorsslaer, Wagstaff, and Lindelöw) and I am using
> the following commands:
>
> factor $assets if rural==1, pcf factors(1)
> predict asset_index if rural==1
> sum asset_index
> xtile quintile=asset_index if rural==1 [aw=hhweight], nq(5)
>
> I am developing the asset index using MICS household survey data to apply to
> a client exit interview where asset data was collected.  The problem that I
> am running into is that when I calculate the asset score for a household by
> hand I get different maximum and minimum asset scores than what stata
> provides.  I am using the formula provided in the WB bank, to sum the scores
> for each asset for each household.  The score for an asset when it is
> present is equal to 1 minus the mean, divided by the SD, times the factor
> loading.  The score when the asset is absent is equal to 0 minus the mean,
> divided by the SD, times the factor loading.
>
> Can anyone tell me if stata uses the same formula or if it adjusts the
> formula when predicting an asset index?  I am trying to figure out why I get
> VERY different scores from the hand calculation.
>
> Thank you very much in advance!
>
> Jennifer

*
*   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/