To use the psc package you need to have two things:
A predictive statistical model in the form of a flexible parametric model (that can be used to predict the outcome of patients had they been treated with the control treatment).
A data cohort that were treated with the experimental treatment.
It is important that your data is formatted correctly before using the psc package.
Let’s load in our data and view it…
load("espac4gemcap.R")
head(espac4_gemcap) ResecM LymphN Diff_Status treat time cen PostOpCA199
2 0 1 2 GEMCAP 8.707019 1 1.383791
5 0 0 0 GEMCAP 49.277267 1 0.000000
8 0 1 1 GEMCAP 6.735929 1 1.824549
9 1 0 1 GEMCAP 21.911958 1 1.131402
10 1 1 0 GEMCAP 23.554534 1 5.855931
12 1 0 0 GEMCAP 118.628176 0 2.595255We should also load in the model and view the data that was used to build the model…
load("flsm.R")
head(flsm$data$m) Surv(time, cen) LymphN ResecM Diff_Status PostOpCA199 (weights)
18 11.300920 1 0 0 6.222576 1
19 14.816032 1 0 1 4.394449 1
28 17.181340 1 0 2 9.314430 1
40 9.789750 1 0 0 4.077537 1
47 5.256242 1 1 0 2.772589 1
52 23.028909 1 0 1 8.398860 1Note that the time and event variables are encoded as ‘time’ and ‘cen’ in both the model and the data cohort. This is key for when we want to make our comparison.
We should also check that the categorical variables have class factor and are dummy encoded, and that continuous variables are class numeric in the data cohort.
str(espac4_gemcap)'data.frame': 362 obs. of 7 variables:
$ ResecM : Factor w/ 2 levels "0","1": 1 1 1 2 2 2 2 1 2 2 ...
$ LymphN : Factor w/ 2 levels "0","1": 2 1 2 1 2 1 2 1 1 1 ...
$ Diff_Status: Factor w/ 3 levels "0","1","2": 3 1 2 2 1 1 1 2 3 2 ...
$ treat : chr "GEMCAP" "GEMCAP" "GEMCAP" "GEMCAP" ...
$ time : num 8.71 49.28 6.74 21.91 23.55 ...
$ cen : int 1 1 1 1 1 0 0 1 0 0 ...
$ PostOpCA199: num 1.38 0 1.82 1.13 5.86 ...If variables are not the correct class this can be changed:
espac4_gemcap$LymphN <- as.factor(espac4_gemcap$LymphN)
espac4_gemcap$PostOpCA199 <- as.numeric(espac4_gemcap$PostOpCA199)It is also important to ensure that the variables present in the model are also present in the comparison dataset, and that their names match exactly:
names(flsm$data$m[,2:5]) %in% names(espac4_gemcap)[1] TRUE TRUE TRUE TRUENow that we have checked the correct variables are in our dataset and that they are encoded correctly, we are ready to use the psc package to make a comparison of two treatments!
Ensure you have a statistical model and a data cohort.
Make sure variables are correctly encoded
Variable names should be consistent in the model and the data cohort