Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: Trend Regression Coefficients of Ordinal Predictor


From   Nick Cox <[email protected]>
To   [email protected]
Subject   Re: st: Trend Regression Coefficients of Ordinal Predictor
Date   Tue, 15 Jan 2013 15:27:18 +0000

I suspect that you are conflating various quite different procedures.

The table you cite certainly doesn't report regression across
quintiles. It seems more likely that one of the tests discussed in the
following FAQ is being used.

FAQ     . . . . . . . . . . . . . .  A comparison of different tests for trend
        . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . W. Sribney
        3/96    Does Stata provide a test for trend?
http://www.stata.com/support/faqs/statistics/test-for-trend/

Whether that's true or not, there are many ways of attaching a P-value
to a trend and not just one, as you seem to imply.

If you have a continuous predictor, it would seem best in most
circumstances to treat it as such. Chopping it into quintile
categories is just throwing away information.

Nick

On Tue, Jan 15, 2013 at 3:15 PM, Karman Tandon <[email protected]> wrote:

> In many papers, a continuous variable will be regressed across
> quintiles of another variable, resulting in a beta coefficient for
> each level of the predictor variable. Then, the authors will write a
> "p for trend" to show that the beta coefficients are significantly
> trending as the levels of the predictor variable increase or decrease.
> What does "p for trend" mean statistically, and how can I arrive at a
> "p for trend" using Stata?
>
> Here is an example of the above in a table from a journal article in
> Cancer Epidemiology, that anyone should be able to access:
> http://cebp.aacrjournals.org/content/14/12/2881/T3.expansion.html

On Jan 14, 2013, at 21:45, karmantandon <[email protected]> wrote:

>> > I am fitting a regression model with a normally distributed outcome (y)
>> > and an ordinal variable in 4 levels (x) as predictor, along with other
>> > variables (a, b), using the command
>> >
>> > regress y i.x a b
>> >
>> > I get separate coefficients for the levels of the predictor x while the
>> > first level is treated as reference. I would like to demonstrate that the
>> > coefficients of the ordinal variable levels are trending in a certain
>> > direction with the variable itself. Is there a way to demonstrate this?
*
*   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index