Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

st: Confusing collinearity error message from -ice-


From   Mike Lacy <[email protected]>
To   [email protected]
Subject   st: Confusing collinearity error message from -ice-
Date   Tue, 30 Nov 2010 21:11:23 -0700


I'm using -ice- in order to take advantage of its -interval() command, so as to be able to prevent nonnegative variables (e.g., area of land owned) from having negative values imputed. I keep getting a collinearity error message whenever I try to run it while including the variables specifying the lower and upper limits, as -interval- requires. I'm using -ice- 1.7.3. Here's what I have for code and relevant output:

// Create lower and upper limits variables for -interval- option
. gen llagsale = 0
. gen ulagsale = 500 // max observed
. gen llacres = 0
. gen ulacres = 1300  // max observed
. gen llincpert = 0
. gen ulincpert = 100 // max observed
. //
. ice years female workoff  ///
> agsales1000 acres100 incpert llagsale ulagsale llacres ulacres llincpert ulincpert ///
>   spiritualconnect preservation econdependence,     m(10) clear  ///
> interval(agsales1000:llagsale ulagsale, acres100:llacres ulacres, incpert:llincpert ulincpert)

   ... snip, snip

   Variable | Command | Prediction equation
------------+---------+-------------------------------------------------------
   .snip, snip
            |         | spiritualconnect preservation econdependence
    workoff | logit   | years female agsales1000 acres100 incpert

snip, snip
// Error messages start here

Imputing
[Note: in regression for workoff, dropping agsales1000
 acres100 incpert due to collinearity]

[Note: in regression for years, dropping agsales1000 acres100 incpert
due to collinearity]

etc.
============================

What confuses me about the collinearity reports is that the three variables reported as offending are only modestly correlated (0.2-0.3). The problem seems to come from the limit variables, since when I omit them and omit the -interval- option, I don't get the collinearity errors. (The imputed values, though, are way out of range.) So, I would like to use the -interval- option. Any thoughts here about what's going on?

Regards,


*
*   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index