seqerulesdisc: Extract association rules using discrete time regression...
In TraMineRextras: TraMineR Extension

seqerulesdisc

R Documentation

Extract association rules using discrete time regression models

Description

Extract association rules from an object created by the createdatadiscrete function, using discrete time regression models to assess the significance of the extracted rules.

Usage

seqerulesdisc(fsubseq, datadiscr, tsef, pvalue=0.1, supvars=NULL, 
   adjust=TRUE, topt=FALSE, link="cloglog", dep=NULL)

Arguments

`fsubseq`	an object created using the `seqefsub` function and that contains the list of subsequences to be tested for an association
`datadiscr`	the object created by the `createdatadiscrete` function and that contains the person-period data
`tsef`	the data frame containing the original time-to-event dataset (equivalent to the `data` argument of the `createdatadiscrete` function)
`pvalue`	the default threshold p-value to consider an association rule as significative, default is 0.1
`supvars`	a vector of variable names to be used as control variables in the regression models (experimental)
`adjust`	if set to TRUE, a Bonferroni adjustment is applied to the p-value threshold specified in the `pvalue` argument
`topt`	if set to TRUE, use an alternative algorithm to extract the rules (very experimental) ; default to FALSE
`link`	the link function to be used in the generalized linear regression model. To obtain hazard ratios, use the complementary log-log link function ("cloglog", as default). The other choice is to use a logit link function ("logit").
`dep`	if set to NULL, test all possible association rules. If an event is set, the function will only test association rules ending with this event

Details

This function uses a list of subsequences created by the seqefsub function from the TraMineR package and tests each possible association rules. It then shows the association rules whose significance, assessed using a discrete time regression model, is higher than the specified p-value threshold.

The algorithm is described in the Müller et al. (2010) article, even though this function uses a discrete time regression model instead of the Cox regression model described in the article. A more complete explanation of the method is available in Müller (2011).

Value

a list with one person-period data frame by event, where the dependent event is different each time. Please see the attached data file and code for an example.

Author(s)

Nicolas S. Müller

References

Müller, N.S., M. Studer, G. Ritschard et A. Gabadinho (2010), Extraction de règles d'association séquentielle à l'aide de modèles semi-paramétriques à risques proportionnels, Revue des Nouvelles Technologies de l'Information, Vol. E-19, EGC 2010, pp. 25-36.

Müller, N.S. (2011), Inégalités sociales et effets cumulés au cours de la vie : concepts et méthodes, Thèse de doctorat, Faculté des sciences économiques et sociales, Université de Genève, http://archive-ouverte.unige.ch/unige:17746.