Bookmark and Share

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


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

Re: st: types and codes of the non-linear models


From   Nick Cox <njcoxstata@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: types and codes of the non-linear models
Date   Thu, 21 Feb 2013 14:27:17 +0000

There is unfortunately no point in trying to interpret the result of
this fit. The intercept has exploded to 3.9 x 10^116 and no standard
error can be calculated for it. Conversely, the coefficient has a
suspiciously narrow c.i. of [.8763176,.8763978]. I'll wager that
nothing in epidemiology is that certain, other than the probability
that we all die sometime.

If you plot this fit against your data, my bet is that this will leap
out at you as a lousy fit. That doesn't necessarily mean that the
model equation is no use for your data, but it does necessarily mean
that this particular fit is useless. However, I would suspect some
basic error in your data, or minimally an outlier. Again, plotting
your data should clarify this.

Nick

On Thu, Feb 21, 2013 at 1:59 PM, BASSILI, Dr Amal     STB/TDR
<bassilia@emro.who.int> wrote:

> I have done the below non-linear regression to predict incidence of a disease over years and would like to know how to interpret the trend. If b2 =0.87, does this mean that the average trend is 0.87% per year?
>
>
> -----------------------------
> nl exp2 : incidence_rate year
> (obs = 7)
>
> Iteration 0:  residual SS =  66.15485
> Iteration 1:  residual SS =  66.15484
>
>       Source |       SS       df       MS
> -------------+------------------------------         Number of obs =         7
>        Model |  387.559441     0           .         R-squared     =    0.8542
>     Residual |  66.1548444     6  11.0258074         Adj R-squared =    0.8542
> -------------+------------------------------         Root MSE      =  3.320513
>        Total |  453.714286     6  75.6190476         Res. dev.     =  35.58776
>
> 2-parameter exp. growth curve, incidence_rate = b1*b2^year
> ------------------------------------------------------------------------------
> incidence_~e |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>          /b1 |   3.9e+116          .        .       .            .           .
>          /b2 |   .8763577   .0000164 53453.24   0.000     .8763176    .8763978
> ------------------------------------------------------------------------------
>   Parameter b1 taken as constant term in model & ANOVA table

Maarten Buis

> On Wed, Feb 20, 2013 at 8:28 PM, BASSILI, Dr Amal     STB/TDR wrote:
>> Please let me know the types and STATA codes of the non-linear models that can forecast the incidence rate of a disease if linear regression cannot be used.
>
> The number of options open to you is just too large to list here. We could write a book-length post here with lots of options and code.
> However, this would require a lot of work from us (for free), and most of it would be useless to you as it would not apply to your problem.
> In order to get a more useful response you need to narrow your question down by giving us more details on what you want to do.
*
*   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/


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