Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.

# Re: AW: st: prgen option quadratic term

 From Nick Cox To statalist@hsphsun2.harvard.edu Subject Re: AW: st: prgen option quadratic term Date Wed, 7 Nov 2012 11:36:09 +0000

```So, you can see what the fit implies statistically. What is your
question then? If you want help with (socio-)economic interpretation,
even people in your field would need to be told what your outcome or
response variable is. Here it's just -y-.

Nick

On Wed, Nov 7, 2012 at 11:29 AM, Meulemann  Max <mmeulemann@ethz.ch> wrote:

> I had hoped that was taken care of by the margins command
>
> ologit y capitagdp c.capitagdp#c.capitagdp i.socioecovars
> margins, at(c_capitagdp=(0.5(0.5)9))  atmeans predict(outcome(4))
> marginsplot
>
> which looked the same as
>
> ologit y c_logcapitagdp c_logcapitagdpsquared \$socioecovars
> oprobpr c_logcapitagdp, levels(c_logcapitagdpsquared=c_logcapitagdp^2) cat(4) newobs(100)
>
> In the linear model that graph is downward slooping and in the quadratic model it is downward slooping and then turning up
>
> ________________________________________
> Von: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu]&quot; im Auftrag von &quot;Nick Cox [njcoxstata@gmail.com]

> When you are fitting a linear term and a quadratic term, the two act
> together. It's their combined effect that you need to interpret. This
> is usually easier with a graph. For example, one negative coefficient
> and one positive coefficient mean a turning point somewhere, but not
> necessarily within the range of the data.
>
> Nick
>
> On Wed, Nov 7, 2012 at 10:35 AM, Meulemann  Max <mmeulemann@ethz.ch> wrote:
>> thx, the margins and marginsplot function helped a lot.
>>
>> Anyways anybody got a idea how I am to interpret my data here: I find a negativ linear term for capitagdp, but if i include a quadratic gdp term, that one turns out to be positiv and significant with a lrtest.
>> Plotting with marginsplot tells me then that for the linear model the probability to agree with a certain issue decreases whereas it increases with capitagdp for the quadratic model, keeping fixed my other covariates

>> Von: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu]&quot; im Auftrag von &quot;Maarten Buis [maartenlbuis@gmail.com]