Re: st: Twoway quadratic prediction plots with CIs in a multivariate regression setting

[Scott Merryman <scott.merryman@gmail.com>]

The -margins- command has a post option to save the results.  The
example below using Roger Newson's -xsvmat- (ssc desc xsvmat)

webuse nhanes2.dta,clear
qui reg iron c.age##c.age i.race zinc
qui margins , at(age = (10(10)80)  (base) race (zero) zinc)  post
mat a= r(table)
mat A = a'

qui reg iron c.age##c.age i.race zinc
qui margins , at(age = (10(10)80)  ) post
mat b = r(table)
mat B = b'
mat c = r(at)
mat atage = c[1..8,1]

xsvmat A,  name(col) saving(cond.dta,replace)
xsvmat B,  name(col)  saving(uncond.dta,replace)
xsvmat atage,  name(col) saving(age.dta,replace)
use b ll ul using cond,clear
rename * =_cond
merge 1:1 _n  using uncond, nogenerate noreport keepusing(b ll ul)
keep b* ll* ul*
merge 1:1 _n  using  age, nogenerate noreport

line b* age, lc(red blue) ///
 || rcap ul_cond ll_cond age, lc(red) ///
 || rcap ul ll age , lc(blue) ///
 ||, legend(order(1 "Conditional" 2 "Unconditional"))


Scott


On Thu, Jul 26, 2012 at 3:51 AM, Michael Ingenhaag
<Michael.Ingenhaag@unil.ch> wrote:
> Thank you Scott.
>
> I was using something similar, but instead of your proposition
>
> The problem with the marginsplot command is, that I cannot combine other
> graphs as I would like to put graphs of two different models in one plot,
> i.e. for an unconditional and a conditional model. Hence, I was thinking
> about qfitci.
>
> Any further suggestions?
*
*   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/