[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: spline

From   Maarten buis <>
Subject   Re: st: spline
Date   Wed, 6 Feb 2008 14:39:51 +0000 (GMT)

--- "A. Manzoni" <> wrote:
>  Here is my case: I want to model recall bias in retrospective
> surveys, looking at the length of the recall period (i.e. distance
> between the events reconstructed and the interview) and I was
> suggested to try to use a spline model.
> I have some theories concerning the shape of the recall (i.e. for
> example I may think that from 10 years onwards it doesn't make any
> difference how much the time distance to the interview is), but I
> know it should be somehow possible to get the number and the position
> of knots which better describe what's going on, besides the shape of
> the effect.

With splines there is a number of issues:
o the number of knots
o the location of the knots
o is the curve between knot linear or quadratic or cubic or...

A simple representation of your theory would be a linear spline with a
single knot at 10 years. You would make these variables like this:

mkspline yr1 10 yr2 = year

If your theory is true than you would expect that the coefficient of
yr2 is 0 

An elaboration would be that you want to estimate where that knot is.
This was discussed in this post:

It may be that you just want to allow the effect of time to have some
flexible non-linear effect. In that case you could have a look at
Patrick Royston's -mvrs- command (see: -findit mvrs-), which is
described in (Royston and Sauerbrei 2007).

Hope this helps,

Patrick Royston and Willi Sauerbrei (2007) "Multivariable modeling with
cubic regression splines: A principled approach". The Stata Journal, 
7(1): 45-70.

Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

Sent from Yahoo! Mail - a smarter inbox

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index