Problems with Groups in Traminer

Question

I'm having a problem using the group function in TraMineR. I have a data set that contains SPELL data, so multiple rows per case. I also have demographic data per case, at one row per case. I merge these together and end up with data that has a demographic covariate per row, so multiple rows per case. An example of this data would be:

id startmin stopmin activity educ4
4      1      20       work    HS
4      20     40       play    HS
8      1      15       sleep   College
8      15     40       work    College

I can make sequence data from this, but when I try to run a plot using the group command

      seqiplot(atus.seq, group = atus.seqdata$educ4, border=NA, 
               withlegend="right",  cex.plot=.55)

it tells me:

    >"Error: group must contain one value for each row in the sequence object".

I have gotten this to work with the example of the mvad data in the training manual, but I can't seem to get it to work with the groups, whether I link to the original demographic data, the merged data, or try to pass the covariates by seqformat and seqdef. Ideas?

Answer 1

The problem comes from your atus.seqdata$educ4 variable which is of length 4 while you have only two sequences. This is the normal result if you merge a table with one row per case with a table with multiple rows per case such as spell data.

When there are covariates in the SPELL file, those covariates should be aggregated into a table with one value per id. The following example with SPELL data including a covariate sex, demonstrates how this can be done.

library(TraMineR)

## Small example of SPELL data
id    <- c(1,1,2,2,3,3,4,4)
start <- c(20,23,20,25,20,21,20,22)
end   <- c(23,26,25,26,21,26,22,26)
state <- c("A","B","A","B","A","B","A","B")
sex   <- c("M","M","F","F","F","F","M","M")

da.spell <- data.frame(id,start,end,state,sex)

## Transform sequences into STS 

da.sts <- seqformat(da.spell, id="id", begin="start", end="end", 
       status="state", from="SPELL", to="STS", process=FALSE)

## Aggregate sex covariate

da.v <- da.spell[,c("id","sex")]
## transform sex factor into a numeric variable for aggregation
sex.lev <- levels(da.v$sex) ## first save sex levels
    da.v$sex <- as.numeric(da.v$sex)
    ## now aggregate table of covariates
    da.av <- aggregate(da.v, by = list(da.v$id), FUN= min)[,-1]
## set aggregated sex as factor with original levels
da.av$sex <- factor(da.av$sex)
levels(da.av$sex) <- sex.lev

## Now we can use TraMineR functions

da.seq <- seqdef(da.sts, cnames=20:26)
seqiplot(da.seq, group=da.av$sex)

You can indeed put STS data and covariates in a same data.frame if you prefer

da.stsv <- cbind(da.sts, da.av)
da.stsv <- da.stsv[,-which(names(da.stsv)=="id")] ## deleting "id" column