Hi!
I have an issue which I previously have received great help on in this and this thread. Unfortunately, a new issue recently came across. In short: I have a follow-up dataset of patients after surgery. I want to investigate wether or not they develop malignancy after having had surgery. Patients are followed six months post-op, thereafter annually. Since there are no date of diagnosis, I have used this code below to randomize a date of diagnosis between last follow-up where there was no diagnosis and the follow-up where malignancy diagnosis was noted.
I had an issue with negative values, which I thought was because malignancy was noted on the first follow up post surgery, and received advice on how to change this (see below). However, now, I have identified that some cases where that is the case and malignancy is noted during the first follow up post surgery (e.g. there is no cum_malignancy == 0), it still produces a value which thereafter leads to the variable time_to_malignancy becoming negative, since the date variable wanted states it developed a malignancy pre surgery (in this patient selection, that is not possible).
The code I am using:
Dataex:
In this example, the only one behaving correctly is B5050. B1010 I can manage with my code replacing all num_days<0, but B4040 just does not make sense.
Any ideas will be greatly appreciated!
I have an issue which I previously have received great help on in this and this thread. Unfortunately, a new issue recently came across. In short: I have a follow-up dataset of patients after surgery. I want to investigate wether or not they develop malignancy after having had surgery. Patients are followed six months post-op, thereafter annually. Since there are no date of diagnosis, I have used this code below to randomize a date of diagnosis between last follow-up where there was no diagnosis and the follow-up where malignancy diagnosis was noted.
I had an issue with negative values, which I thought was because malignancy was noted on the first follow up post surgery, and received advice on how to change this (see below). However, now, I have identified that some cases where that is the case and malignancy is noted during the first follow up post surgery (e.g. there is no cum_malignancy == 0), it still produces a value which thereafter leads to the variable time_to_malignancy becoming negative, since the date variable wanted states it developed a malignancy pre surgery (in this patient selection, that is not possible).
The code I am using:
Code:
bys id_code (surg_date followup_date): gen cum_malignancy = sum(malignancy) gen num_days = (followup_date-followup_date[_n-1])-1 if cum_malignancy == 1 & cum_malignancy[_n-1] == 0 set seed 123 gen wanted = followup_date[_n-1] + runiformint(1,num_days) if cum_malignancy == 1 & cum_malignancy[_n-1] == 0 replace num_days = followup_date - surg_date if num_days<0 replace wanted = surg_date + runiformint(1,num_days) if !missing(num_day) & missing(wanted) format %td wanted drop if !(cum_malignancy == 1 & cum_malignancy[_n-1] == 0) gen time=datediff(surg_date, wanted, "day" > )
Code:
* Example generated by -dataex-. For more info, type help dataex clear input str16 id_code float malignancy long surg_date float(followup_date cum_malignancy num_days wanted) "B1010" 1 15910 16310 1 -5440 . "B1010" 1 15910 16642 2 . . "B1010" 1 15910 17027 3 . . "B1010" 1 15910 17372 4 . . "B1010" 1 15910 17741 5 . . "B1010" 1 15910 18105 6 . . "B1010" 1 15910 18469 7 . . "B1010" 1 15910 18861 8 . . "B2020" 0 15918 16342 0 . . "B2020" 0 15918 16799 0 . . "B2020" 0 15918 17074 0 . . "B2020" 0 15918 17499 0 . . "B2020" 1 15918 17822 1 322 17610 . "B3030" 1 21067 21174 1 3780 20739 "B4040" 1 17347 17714 1 1634 16758 "B4040" 1 17347 18079 2 . . "B4040" 1 17347 18449 3 . . "B5050" 0 15103 15526 0 . . "B5050" 0 15103 15865 0 . . "B5050" 0 15103 16199 0 . . "B5050" 0 15103 16569 0 . . "B5050" 0 15103 16931 0 . . "B5050" 0 15103 17304 0 . . "B5050" 0 15103 17660 0 . . "B5050" 1 15103 18025 1 364 17778 end format %td surg_date format %td followup_date format %td wanted
Any ideas will be greatly appreciated!
