Using R for the self-controlled case series method

There are two ways in which we can fit the case series model after the data have been reformatted:

  1. Download the gnm package and use gnm to fit a conditional poisson regression model with eliminate = indiv (where indiv is a factor for each individual in the data set)
  2. Use the survival package (included in R 2.5.0, so no need to download) and use clogit to fit a logistic regression model with strata = event (where event is a factor for each event in the data set, rather than a factor for each individual). Here, we fit a conditional logistic regression model rather than a conditional poisson regression model: because recurrent events are assumed to be independent the conditional logistic likelihood is equivalent if each event is treated like a separate individual.

All examples on this webpage and the R package use clogit, therefore reformatted data to fit SCCS models looks a little different in R than in other packages.  

R package

A general R package for self-controlled case series models was written by Yonas Weldesselassie. This can fit:

  1. The standard SCCS model.
  2. The semiparametric SCCS model. (Farrington CP and Whitaker HJ. Semi-parametric analysis of case series data. JRSS C, 2006, 55(5): 553-594.)
  3. The SCCS model for event dependent observation periods. (Farrington CP et al. Self-Controlled Case Series Analysis with Event-Dependent Observation Periods. JASA, 2011, 106(494): 417–426.)
  4. The SCCS model with smooth (spline-based) age effect. (Weldeselassie YG, Whitaker HJ, Farrington CP. Self-Controlled Case Series Method with Smooth Age Effect. Stats in Med, 2014, 33(4): 639-649.)
  5. The SCCS model with smooth (spline-based) exposure effect, for a single exposure. (Weldeselassie YG, Whitaker HJ, Farrington CP. Flexible modelling of vaccine effect in self-controlled case series models. Biometrical journal, early view available)
  6. The non-parametric, spline based SCCS model, for a single exposure. (Weldeselassie YG, Whitaker HJ, Farrington CP. Non-parametric self-controlled case series method. Submitted. See Yonas' thesis.)
It also includes the three data sets used in the tutorial paper, a function for simulating data suitable for SCCS analyses and a function to reformat the data, ready to fit the model using clogit.

We plan to extend this package in the future to fit further SCCS models.

SCCS package version 1.0

SCCS package manual

R package for the case series analysis for censored, perturbed or curtailed post event exposures

An R package, adSCCS, has been written by Ronny Kuhnert (Robert Koch Institute, Berlin) to fit the extended self-controlled case series method for the situation when no exposure can occur after an event. This is the method outlined in the paper: Farrington, Whitaker and Hocine (2008). Case series analysis for censored, perturbed, or curtailed post-event exposures. Biostatistics, 10(1): 3-16.

adSCCS package version 1.5

adSCCS manual

Simple examples for the tutorial paper

All script files were written by Heather Whitaker for R version 2.5.0 (free software available from http://www.r-project.org/). Please let us know if you have any suggestions for improving them.

MMR and meningitis in Oxford example

To run the MMR and meningitis in Oxford example detailed in the tutorial paper save these two files:

'ox.txt', the data in a tab-delimited text file.

'oxford.r', the commands in an r script file.

Open oxford.r and select 'run all' under the edit menu.

ITP and MMR example

'itp.r' fits the multiple risk periods example on p.1782-1783 of the tutorial paper.

'itp.txt', data

'itp.r', r script file

Intussusception and oral polio vaccine example

'intuss.r' fits analysis 5, repeat exposures example detailed on p.1787-1789 of the tutorial paper.

'intuss.txt', data

'intuss.r', r script file 

The self-controlled case series method / Heather Whitaker / updated December 2015