Problem4

#Setup for P4

rm(list=ls())
library("survival")
df1<-read.csv("Breast_cancer.csv", header=TRUE)
head(df1)
str(df1)
## 'data.frame':    45 obs. of  4 variables:
##  $ id   : int  1 2 3 4 5 6 7 8 9 10 ...
##  $ times: int  224 212 208 198 181 148 101 100 71 70 ...
##  $ delta: int  0 0 0 0 1 1 0 0 0 0 ...
##  $ Z    : int  0 0 0 0 0 0 0 0 0 0 ...

#(a) Compute the Kaplan-Meier Estimator for S(t)

lsurv1 <- with(df1, Surv(times, delta == 1))
lsurv1
##  [1] 224+ 212+ 208+ 198+ 181  148  101+ 100+  71+  70+  69   47   23    5    8 
## [16]  10   13   18   24   25   26   31   35   40   41   48   50   59   61   68 
## [31]  71   76+ 105+ 107+ 109+ 113  116+ 118  143  154+ 162+ 188+ 212+ 217+ 225+
km1<-survfit(lsurv1~1, data= df1, conf.type = "log-log")
print(km1)
## Call: survfit(formula = lsurv1 ~ 1, data = df1, conf.type = "log-log")
## 
##       n events median 0.95LCL 0.95UCL
## [1,] 45     26    113      50      NA
summary(km1)
## Call: survfit(formula = lsurv1 ~ 1, data = df1, conf.type = "log-log")
## 
##  time n.risk n.event survival std.err lower 95% CI upper 95% CI
##     5     45       1    0.978  0.0220        0.853        0.997
##     8     44       1    0.956  0.0307        0.834        0.989
##    10     43       1    0.933  0.0372        0.807        0.978
##    13     42       1    0.911  0.0424        0.780        0.966
##    18     41       1    0.889  0.0468        0.753        0.952
##    23     40       1    0.867  0.0507        0.727        0.938
##    24     39       1    0.844  0.0540        0.701        0.923
##    25     38       1    0.822  0.0570        0.676        0.907
##    26     37       1    0.800  0.0596        0.651        0.891
##    31     36       1    0.778  0.0620        0.626        0.874
##    35     35       1    0.756  0.0641        0.602        0.856
##    40     34       1    0.733  0.0659        0.578        0.839
##    41     33       1    0.711  0.0676        0.555        0.821
##    47     32       1    0.689  0.0690        0.532        0.802
##    48     31       1    0.667  0.0703        0.509        0.784
##    50     30       1    0.644  0.0714        0.487        0.765
##    59     29       1    0.622  0.0723        0.465        0.746
##    61     28       1    0.600  0.0730        0.443        0.726
##    68     27       1    0.578  0.0736        0.421        0.706
##    69     26       1    0.556  0.0741        0.400        0.686
##    71     24       1    0.532  0.0745        0.378        0.665
##   113     16       1    0.499  0.0769        0.342        0.638
##   118     14       1    0.463  0.0793        0.304        0.608
##   143     13       1    0.428  0.0808        0.269        0.578
##   148     12       1    0.392  0.0815        0.235        0.546
##   181      9       1    0.349  0.0833        0.193        0.509

#(b) Compute the 95% confidence intervals for sK-M(t) assuming conf.type=“plain”. What an issue?

survfit(lsurv1~1, data= df1, conf.type = "plain")
## Call: survfit(formula = lsurv1 ~ 1, data = df1, conf.type = "plain")
## 
##       n events median 0.95LCL 0.95UCL
## [1,] 45     26    113      59      NA
print(survfit(lsurv1~1, data= df1, conf.type = "plain"))
## Call: survfit(formula = lsurv1 ~ 1, data = df1, conf.type = "plain")
## 
##       n events median 0.95LCL 0.95UCL
## [1,] 45     26    113      59      NA

In some cases, the confidence limits calculated this way would extend below 0 or above 1.

#(c) Compute the 95% confidence intervals assuming conf.type=“log”. What an issue?

survfit(lsurv1~1, data= df1, conf.type = "log")
## Call: survfit(formula = lsurv1 ~ 1, data = df1, conf.type = "log")
## 
##       n events median 0.95LCL 0.95UCL
## [1,] 45     26    113      59      NA
print(survfit(lsurv1~1, data= df1, conf.type = "log"))
## Call: survfit(formula = lsurv1 ~ 1, data = df1, conf.type = "log")
## 
##       n events median 0.95LCL 0.95UCL
## [1,] 45     26    113      59      NA

By taking log and adding and subtracting an error term, the bounds of our interval can lie outside [0,1]

#(d) Compute the 95% confidence intervals assuming conf.type=“log-log”

survfit(lsurv1~1, data= df1, conf.type = "log-log")
## Call: survfit(formula = lsurv1 ~ 1, data = df1, conf.type = "log-log")
## 
##       n events median 0.95LCL 0.95UCL
## [1,] 45     26    113      50      NA
print(survfit(lsurv1~1, data= df1, conf.type = "log-log"))
## Call: survfit(formula = lsurv1 ~ 1, data = df1, conf.type = "log-log")
## 
##       n events median 0.95LCL 0.95UCL
## [1,] 45     26    113      50      NA

#(e) Compute Nelson-Aalen estimator for the cumulative hazard(t)

naest1 <- cumsum(km1$n.event/km1$n.risk)
naest1
##  [1] 0.02222222 0.04494949 0.06820531 0.09201483 0.11640508 0.14140508
##  [7] 0.16704610 0.19336189 0.22038892 0.24816670 0.27673813 0.30614989
## [13] 0.33645292 0.36770292 0.39996098 0.43329432 0.46777708 0.50349136
## [19] 0.54052840 0.57898994 0.57898994 0.62065660 0.62065660 0.62065660
## [25] 0.62065660 0.62065660 0.62065660 0.62065660 0.68315660 0.68315660
## [31] 0.75458518 0.83150825 0.91484159 0.91484159 0.91484159 1.02595270
## [37] 1.02595270 1.02595270 1.02595270 1.02595270 1.02595270 1.02595270
## [43] 1.02595270
plot(km1$time, naest1, type="s", col="red",
     xlab="Time(Days)", ylab = "Cumulative Hazard", main = "Nelson-Aalan Estimator")

#(f) Compute the Fleming-Harrington estimator for S(t). Comment on the results compared to (a)

fh1 <- exp(-naest1)
fh2 <- survfit(lsurv1 ~ 1, data = df1, type = "fleming-harrington")
plot(fh2, xlab= "Time(Days)", ylab = "Survival Probability",
     main = "Fleming-Harrington Estimator", col="orange",
     conf.int = F)
points(km1$time, fh1, type = "s", col="red")
points(km1$time, km1$surv, type = "s", col="blue")

legend("topright", legend = c("F-H 1", "F-H 2", "K-M"),
       col = c("red","orange","blue"), lty=1)

cbind(fh1, fh2$surv, km1$surv)
##             fh1                    
##  [1,] 0.9780229 0.9780229 0.9777778
##  [2,] 0.9560458 0.9560458 0.9555556
##  [3,] 0.9340687 0.9340687 0.9333333
##  [4,] 0.9120916 0.9120916 0.9111111
##  [5,] 0.8901146 0.8901146 0.8888889
##  [6,] 0.8681376 0.8681376 0.8666667
##  [7,] 0.8461606 0.8461606 0.8444444
##  [8,] 0.8241837 0.8241837 0.8222222
##  [9,] 0.8022067 0.8022067 0.8000000
## [10,] 0.7802299 0.7802299 0.7777778
## [11,] 0.7582530 0.7582530 0.7555556
## [12,] 0.7362763 0.7362763 0.7333333
## [13,] 0.7142995 0.7142995 0.7111111
## [14,] 0.6923228 0.6923228 0.6888889
## [15,] 0.6703462 0.6703462 0.6666667
## [16,] 0.6483696 0.6483696 0.6444444
## [17,] 0.6263931 0.6263931 0.6222222
## [18,] 0.6044167 0.6044167 0.6000000
## [19,] 0.5824404 0.5824404 0.5777778
## [20,] 0.5604642 0.5604642 0.5555556
## [21,] 0.5604642 0.5604642 0.5555556
## [22,] 0.5375913 0.5375913 0.5324074
## [23,] 0.5375913 0.5375913 0.5324074
## [24,] 0.5375913 0.5375913 0.5324074
## [25,] 0.5375913 0.5375913 0.5324074
## [26,] 0.5375913 0.5375913 0.5324074
## [27,] 0.5375913 0.5375913 0.5324074
## [28,] 0.5375913 0.5375913 0.5324074
## [29,] 0.5050203 0.5050203 0.4991319
## [30,] 0.5050203 0.5050203 0.4991319
## [31,] 0.4702056 0.4702056 0.4634797
## [32,] 0.4353921 0.4353921 0.4278274
## [33,] 0.4005801 0.4005801 0.3921751
## [34,] 0.4005801 0.4005801 0.3921751
## [35,] 0.4005801 0.4005801 0.3921751
## [36,] 0.3584548 0.3584548 0.3486001
## [37,] 0.3584548 0.3584548 0.3486001
## [38,] 0.3584548 0.3584548 0.3486001
## [39,] 0.3584548 0.3584548 0.3486001
## [40,] 0.3584548 0.3584548 0.3486001
## [41,] 0.3584548 0.3584548 0.3486001
## [42,] 0.3584548 0.3584548 0.3486001
## [43,] 0.3584548 0.3584548 0.3486001

The F-H estimator that is more sensitive to later events, it would tend to show a lower survival probability later compared to the K-M estimator.

##Problem5 #Setup

library(haven)
df2<-file.choose() #choose anemia2.sas7dbat
df2<-read_sas(df2)
df2$ageI <- as.factor(ifelse(df2$age >= 20,1,0))
km2.byage<- survfit(Surv(obs_time, GVHD==1)~ageI, data=df2, conf.type="log-log")
km2.byage
## Call: survfit(formula = Surv(obs_time, GVHD == 1) ~ ageI, data = df2, 
##     conf.type = "log-log")
## 
##         n events median 0.95LCL 0.95UCL
## ageI=0 28      5     NA      NA      NA
## ageI=1 36     15     NA      25      NA
summary(km2.byage)
## Call: survfit(formula = Surv(obs_time, GVHD == 1) ~ ageI, data = df2, 
##     conf.type = "log-log")
## 
##                 ageI=0 
##  time n.risk n.event survival std.err lower 95% CI upper 95% CI
##    10     28       1    0.964  0.0351        0.772        0.995
##    22     27       2    0.893  0.0585        0.704        0.964
##    28     25       1    0.857  0.0661        0.663        0.944
##    49     24       1    0.821  0.0724        0.623        0.921
## 
##                 ageI=1 
##  time n.risk n.event survival std.err lower 95% CI upper 95% CI
##     8     35       1    0.971  0.0282        0.814        0.996
##     9     34       1    0.943  0.0392        0.790        0.985
##    11     33       1    0.914  0.0473        0.757        0.972
##    12     32       1    0.886  0.0538        0.724        0.955
##    16     30       1    0.856  0.0595        0.688        0.938
##    17     29       1    0.827  0.0644        0.654        0.918
##    20     28       2    0.768  0.0721        0.588        0.876
##    25     26       3    0.679  0.0799        0.496        0.808
##    28     22       1    0.648  0.0820        0.464        0.783
##    31     21       1    0.617  0.0837        0.433        0.757
##    35     20       1    0.586  0.0850        0.403        0.731
##    46     19       1    0.556  0.0859        0.374        0.704