Re: st: Tobit graphics

[Nick Cox <njcoxstata@gmail.com>]

-yaxis(1 2)- is a standard -twoway- option. It belongs in your
-addplot()- option.

Nick

On Sun, Dec 30, 2012 at 1:53 PM, Kakouros, Nikolaos
<Nikolaos.Kakouros@umassmemorial.org> wrote:
> Where do you declare yaxis(1 2)  though?
>  It is not an option for marginsplot and setting it in the first addplotted scatter as an option doesn't seem to make a difference.  I used yaxis(2) for the secondary scatter and lfitci but that places the second y axis to the left of the first one.
> Sure - I can move it to the right with the graph editor but where's the fun in that?

Nick Cox [njcoxstata@gmail.com]

> In this kind of problem I usually declare
>
> yaxis(1 2)
>
> and then if stuff I want is in the wrong place so far as I am concerned I write
>
> yaxis(2)
>
> instead of
>
> yaxis(1)
>
> and/or vice versa.
>
> Nick
>
> On Thu, Dec 27, 2012 at 6:23 PM, Kakouros, Nikolaos
> <Nikolaos.Kakouros@umassmemorial.org> wrote:
>
>> The plot thickens on the Tobit graphics.
>>
>> marginsplot, recast(line) recastci(rarea) addplot( (scatter x y) (lfictci x2 y2) (scatter x2 y2))
>>
>> worked, but the second set of data needs a different y axis.
>>
>> I tried
>>
>> marginsplot, recast(line) recastci(rarea) addplot( (scatter x y) (lfictci x2 y2, yaxis(2)) (scatter x2 y2, yaxis(2)))
>>
>> but the second y axis ends up stacked on the left rather than being on the right.
>
> Kakouros, Nikolaos
>
>> Trust me, Dr Cox. For me to get this to work, quite a bit of luck would be needed.
>> Your recommendation worked, though.
>>
>> I used
>> marginsplot, recast(line) recastci(rarea) addplot( (scatter x y) (lfictci x2 y2) (scatter x2 y2))
>>
>> and it worked great!
>
> Nick Cox
>
>> Not clear why luck should be needed, but it's not necessary to specify -twoway-. Otherwise specify what kind of error message you got.
>
> On 27 Dec 2012, at 13:21, "Kakouros, Nikolaos"
> <Nikolaos.Kakouros@umassmemorial.org
>
>>> Ro'I,
>>> Many thanks - that works great!  I changed recastci(rline) to
>>> (rarea) to get the same shaded CI effect as lfitci.
>>>
>>> Now - is it possible to show this data on the same graph as the other
>>> dataset's scatter and lfitci?
>>> I tried
>>> marginsplot, recast(line) recastci(rarea) addplot(twoway (scatter x
>>> y) (lfictci x2 y2) (scatter x2 y2))
>>>
>>> I also tried
>>> Twoway (marginsplot, recast(line) recastci(rarea) addplot(scatter x
>>> y)) (lfictci x2 y2) (scatter x2 y2)
>>>
>>>
>>> with no luck...
>
> Ro'i Zultan
>
>>> I think the following should work for you:
>>> tobit x y, ll(0)
>>> margins, at(y=(0(10)100))
>>> marginsplot, recast(line) recastci(rline) addplot(scatter x y)
>>>
>>> With the margins parameters adjusted for your data.
>
> 2012/12/26 Kakouros, Nikolaos <Nikolaos.Kakouros@umassmemorial.org>
>
>>>> I have a set of data and using a twoway scatter and lfitci to show
>>>> graphically along with the least-squares regression line and
>>>> confidence intervals.
>>>> Another set of data, however, is heavily censored with the dependent
>>>> variable hitting a floor at 0 (and about a third of values actually
>>>> being at zero).
>>>> I noted the recent discussion on Tobit, chaired by Maarten Buis.
>>>>
>>>> I used tobit x y, ll(0)  for the regression but not sure how to
>>>> recreate the lfitci output (regression line and confidence intervals)
>>>> using the tobit regression info to produce the scatter graphic.
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/