生存分析中的生存时间

Question

I'm having problems to analyze a survival dataset that I have. I will put the dput output of the dataset in a github link to not pollute the question.

Here is the data https://gist.github.com/anonymous/4fdff1c6d0853c41939e2a67d9e0e45b

In this dataset, I want to make plot of survival curves for each group, so I need to make a survfit() model.

The variables W1,W2,..,W43 represents weeks and the numbers represents some measure. When I have a dot . in any week, it means that individual died that week, and consequently every week that follows are flagged with dot .

In a survival model this death represents an event (failure) and if the individual survival all the weeks he represents a censored data.

To make a survival model the way that I know I need to have a data like this below

time=c(3,4,8,8,5,2)
event=c(1,1,0,0,1,1)

in this case time represents the time of death in weeks and event is 1 if death and 0 if censored.

EDIT: I thinked in one possible solution, but I don't know how I can do it. The idea is below

1) Take all the columns W1,W2,...,W43 and put 1 if its a number and put 0 if it is a dot .

2) Create a new variable that represents time and the value of this variable will be the sum of columns W1 to W43, so it will W1+W2+...+W43.

3) Create a new variable that represents the event, then if time=43 this means that the individual survived all the time then it will be 0 (censored) and if if is less than 43 it means that the individual died, then the variable will be 1.

Anyone can help me to do it?

Answer 1

I named your dataset sdat and these operations add the two additional columns:

sdat$time= apply(sdat[ ,grepl("W", names(sdat))], 1 , #work by rows on "W"-columns
                    function(r) which( r==".")[1] )  # seq-number of first "."
sdat$event <- as.numeric( !is.na(sdat$time) ) # convert NA's to logical and to 1,0
sdat$time= ifelse( is.na(sdat$time) , 43, sdat$time) # set time to 43 for survivors

 # Check results
 head( sdat[ , !grepl("W", names(sdat))] ) # remove "W" cols
  Group Ref Sex  M1   M2 M3  M4 time event
1    11   4   1 959 1940 10 184   23     1
2    11   4   1 960 1770 10 189   31     1
3    11   4   1 961 1970 10 166   23     1
4    11   4   1 962 1870  1 180   43     0
5    11   4   1 964 1780 11 239   43     0
6    12   4   1 966 1980 11 182   43     1

As an analyst I would be asking what meaning to attach to the varying "W"-numbers leading up to the deaths, but that was not your question.