how to convert PPER format to SPELL format for CoxPH model - survival-analysis

Question

CoxPH Survival Analysis
I have a data set in PPER ( person period) format like :
Machine_id,Timestamp,Event,TDV1,TDV2,TDV3,TDV4
TDV1/2 are factors(brand , location) TDV3/4 are continous(temprature, humidity)
Need to convert to SPELL format like:
Machine_id,start.time,stop.time,event,TDV1,TDV2,TDV3,TDV4
I was able to convert from SPELL to PPER by using seqdef() & toPersonPeriod() in TraMineRextras
Help needed to do the reverse. Also how to treat continous variables while going from PPER to SPELL format?

Answer 1

Person-period data is just a special case of spell data where the start and end times are identical. You get spell data by duplicating your Timestamp variable and renaming the first start.time and the second stop.time.
Possibly, you can aggregate your records using list=c(Machine_id, event, TDV1, TDV2, TDV3, TDV4) as by argument. Proceeding twice, once with FUN="min" and once with FUN="max", you should be able to find the start and end times of the spells with unchanged values of event and the covariates.
I illustrate here with an example
## creating example data
p.df <- data.frame(scan(what=list(Id=0, timestamp=0, event="", work="", income=0)))
1 2000 S working 100
1 2001 S working 100
1 2002 M working 100
1 2003 M working 100
1 2004 M jobless 80
1 2005 M jobless 70
2 2000 S jobless 10
2 2001 S working 100
2 2002 S working 100
## leave previous line blank to end scan
p.df$start <- p.df$timestamp
p.df$end <- p.df$timestamp
p.df <- p.df[,-2] ## deleting timestamp variable
bylist <- list(id = p.df$Id, event=p.df$event,
work=p.df$work, income=p.df$income)
spell1 <- aggregate(p.df[,c("start","end")], by=bylist, FUN="min")
spell2 <- aggregate(p.df[,c("start","end")], by=bylist, FUN="max")
## reordering columns
spell <- spell1[,c(1,5,6,2,3,4)]
spell[,3] <- spell2[,6] ## taking end value from spell2
spell <- spell[order(spell$id,spell$start),] ## sorting rows
spell
## id start end event work income
## 5 1 2000 2001 S working 100
## 4 1 2002 2003 M working 100
## 3 1 2004 2004 M jobless 80
## 2 1 2005 2005 M jobless 70
## 1 2 2000 2000 S jobless 10
## 6 2 2001 2002 S working 100
Hope this helps.