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

# re:st: cubic spline in time series

 From "Ariel Linden, DrPH" To Subject re:st: cubic spline in time series Date Tue, 14 Sep 2010 09:38:12 -0400

```Steve,

Take a look at mvrs (multivariable regression spline models) a user-written
program by Patrick Royston and William Sauerbrei.

This program can create regression splines for one or more variables and
will allow you to choose between step, linear, or cubic spine functions. It
has tremendous flexibility and yet is fully automated.

Ariel

Date: Mon, 13 Sep 2010 14:25:41 -0500
Subject: st: cubic spline in time series

I have number of cases data per week for a multi-week time series.  The data
will be modeled with negative binomial models.

I'd like to account for the general tendencies of the case data over time,
using cubic splines and then construct the negative binomial model including
the cubic splines as dependent variable smoothers, along with time-varying
independent variables to determine the effect of those variables on
deviations from the cubic spline-fitted dependent variable.

I've used the Stata command -mkspline-, which sets spline knots according to
ranges of values of a single variable.  Using -mkspline- with default
arguments with the dependent variable in my model, number of cases, results
in four knots being placed according to increasing values of number of
cases.

I've searched statalist and the manuals for a spline function that takes two
arguments, the variable to be fitted and another variable against which the
fit should be made.  The only thing I see is a Mata command -spline3_, which
takes two arguments, real vector x and real vector y.

This might be what I need but I've never used Mata before.  I wonder if
someone could help with
1. alternative ways to construct a spline fitted time function of a
variable.
2. tell me if Mata -spline3- is what I'm looking for.
3. give me guidance in using either a Mata function or some other code that
could do the spline fit of a variable in time.

Best regards,
Steve Rothenberg
National Institute of Public Health
Mexico

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```