[R] Survival analysis

[Göran Broström]

Dear dr Medic,

Den 2020-04-17 kl. 23:03, skrev Medic:
> On 2020-04-17 20:06, Medic wrote:
>> I can't understand how to do a survival analysis (?Surv ()) when some
>> event occurred before the start of observation (left censored). If I
>> understand correctly, there are two methods. I chose a method with: 1)
>> time from the start of treatment to the event and 2) the indicator of
>> the event. I did (in my data) the event indicator so:
>> 1 - event, 2 - event before the start of observation, 0 - no event
> 
> I have no experience of left censoring beyond the text book.  Is your
> left censored data the SAME event or a different event?
> 
> YES, THE SAME!
> 
>> ---
>> library(survival)
>> left_censor_data <- read.table("left.csv", header = TRUE, sep = ";")
>> #sep = ";" it's right!
>> dput(left_censor_data, file="left_censor_data") #file attached
>> left_censor_data
>>     'data.frame':   11 obs. of  2 variables:
>>     $ timee : int  5 151 33 37 75 14 7 9 1 45 ...
>>     $ eventt: int  2 0 0 0 0 0 0 2 0 1 ...
>>     # 1—event, 2 – event before the start of observation , 0 – no event
> 
> So if I read this data correctly the first observation is left censored.
> What does the time "5" refer to?  Is that 5 days BEFORE observation the
> event happened?
> 
> YES, EXACTLY!
> 
> My text book understanding of left censored data was that your
> censored points would
> have time 0.
> 
> I TRIED TO SET TIME 0 NOW (for censored points), AND RECEIVED THE SAME
> WARNING (AND THE CURVE TURNED OUT WRONG)
> 
>> sur <- Surv(time = left_censor_data$timee,  event =
>> left_censor_data$eventt, type = "left")
>>    WARNING message:
>>    In Surv(time = left_censor_data$timee, event =
>> left_censor_data$eventt,  :
>>    Invalid status value, converted to NA
>>
>> #Why such a WARNING message?

Because the choice "type = 'left'" is incompatible with an event 
variable taking three values, see the help page for survival::Surv. My 
guess from your (vague) description is that you want "type = 
'counting'", as indicated below.

>> #Then everything turns out wrong
> 
> Is the censoring type you want LEFT TRUNCATION rather than LEFT.
> If they are also right censored I think R Surv calls these Counting.
> 
> I SAY ABOUT LEFT CENSORING (NOT ABOUT LEFT TRUNCATION)!
> (COUNTING? I DO NOT UNDERSTAND THIS.)

Then you should read an elementary text on survival analysis or consult 
a local statistician.

G,

> 
> THANKS! I HOPE SOMEONE EXPLAIN TO ME
> 1) HOW TO COMPILE THE DATA and
> 2) WRITE A CODE
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>