Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Wish List Re: Graphs


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: Wish List Re: Graphs
Date   Tue, 19 Nov 2002 09:39:46 -0000

(sorry for previous premature posting)

SamL

> I am sending this to the stata list because I use the Unix
> version in a
> shared environment, and owing to many complexities I have
> not used the
> graphics of stata.  But, I am thinking of buying a laptop
> and graphics
> would be a consideration for statistical software
> purchases.  The laptop
> may run Linux, I'm still thinkin', but the big question
> concerns effort.
> Before I invest great effort in learning how to use the
> graphics in stata,
> maybe some set of people can confirm the wisdom of making
> the investment,
> or warn me to look elsewhere for a solution to some graphic
> aims I have. I
> also hope that this note will be of use to others.
>
> My graphics wishlist:
>
> 1)I want to be able to write down a function (e.g., Y=b0 + b1*X1;
> Pr(y=1)=(e(b0+b1*X1))/(1+e(b0+b1*X1)) ) and have it graphed.

-generate- the variables and then -graph- them using -c(s)-.
Or check out

STB-16  ssi5.1 . . . . . . . . . . . . . . . . . . . . . .  Graphing
functions
        (help fcnplot if installed)  . . . . . . . . . . . . . . . .
W. Gould
        11/93   pp.23--26; STB Reprints Vol 3, pp.188--193
        graph a specified function -- which may be specified by a
        user-written program -- and optionally leaves behind the
        dataset of values of x and f(x) just graphed

>
> 2)I want to be able to write down multiple functions, and
> have them appear
> on the same graph (I will need 6 different functions in my current
> application).

Extension of 1), except that for 6 functions -keyplot- from SSC
will help.

> 3)I want to be able to have a legend saying what lines go with what
> function.

Same as 2).

> 4)I want to be able to provide a series that tells where
> the tick marks
> should be.  Something like (xtick=0,.1, .2, .3, . . . ,
> 1.0).  This would
> be of most use, of course, if the graph started at .27, but
> I still wanted
> tick marks on the .3, .4, and such, not at wierd values.

This is just -xtick()- with your own numlist. Please
see the manual.

>
> 5)I want the graph to have a range for the axes, but I also
> want to be
> able to specify *different* limits for each function.  In
> pictures, using
> two lines as illustration, I want to be able to do
> something like the
> following:
>
> |    /
> |  ./
> |  /.
> |    .
> |     .
> |_________
>
> The dashed line going up (from left to right) does not have
> the same range
> in X as the dotted line (the dotted line is going down from left to
> right).  This is not a feature of the function but is,
> instead, a feature
> of the data used to estimate the model that produced the
> function; the
> range of X is smaller for one group than the other, so I
> don't want the
> line for any group to exceed the range of the data from
> which I estimated
> the function.

Controllable by setting results to missing over
intervals where a function should not be plotted.
>
> 6)The above is great, but it would also be nice to draw the
> line over the
> full range of X, and mark certain points on each line.  The
> figure below
> illustrates what I mean
>
> |.    /
> | -  -
> |  ./
> |  /.
> | /  _
> |-    .
> /______
>
> This would let me have my cake (the horizontal lines
> showing the limits of
> defensible if slightly improper inference) and eat it too
> (the full line,
> showing the logical implications if all groups had the same
> range--useful
> perhaps only for illustration, not inference.
>
> The above could be secured by shading, symbols, or whatever.

-generate- extra variables defining just a few values only.
>
> 7)Multiple lines for titles.

Please see the manual and report back on what can't
be done.

> 8)Easy ability to switch between landscape and portrait
> page orientation.

Check out -sq- from -ssc- and -graph-'s -bbox()- option.
>
> 9)Now, this is really(!) tough, but would be ***great***!
> This is more
> for presentation in a class or among a group of
> non-statisticians.  I'd
> love to estimate a model on some data, then have it draw
> the line showing
> the predicted values on some Y given some (varying) X, at
> means on the
> other Xs (or at prespecified values for categorical data),
> and THEN let me
> re-estimate the model with another X added (X_new)
> (ideally, as suggested
> by the students in the class), and THEN let me draw the old
> X-Y relation
> and another line (dashed, let's say) with the new X-Y
> relation (once the
> X_new) is controlled.  I can think of nothing more able to
> quickly convey
> to students the value of statistical modelling.  I admit
> this is the outer
> limit of my wish list at this time.

Is programmable, no doubt. I am not
aware that anyone has wrapped this up in a program or programs.
-regplot- on SSC might be a starting point.

Nick
n.j.cox@durham.ac.uk

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index