Strange "changeable" results in loop in R - r

Question

Update2
The second set.seed(i) should be replaced as set.seed(i+1), or whatever another new random series. If not, s3 <- sum(data$gene == 0 & data$cancer == 1) will always be 0, since the number smaller than 0.08 will be smaller than 0.39.
I did't correct my original question code because this matter is not related to this post's core question.
Update
set.seed(i) is added twice since there are two random number generations, i.e., random1 and random2. However, the results among operations still changeable, which is strange.
Background:
the codes below is about odds ratios. But the focus point is not statistic here. Instead, I find the results changed (!) among some operations which are actually the same (I suppose that they are in fact not, but I cannot figure it out).
Code:
gene <- vector(length = 500, mode = "integer")
cancer <- vector(length = 500, mode = "integer")
data <- data.frame(gene, cancer)
odd_withMutate <- vector(length = 20, mode = "numeric")
odd_noMutate <- vector(length = 20, mode = "numeric")
result <- data.frame(odd_withMutate, odd_noMutate)
for (i in 1:20) {
# set.seed(12)
# set.seed(16)
set.seed(i)
random1 <- runif(500, min = 0, max = 1)
# set.seed(12)
# set.seed(16)
set.seed(i) # add this instruction
random2 <- runif(500, min = 0, max = 1)
for (j in 1:500) {
if (random1[j] < 0.39){
data[j,1] <- 1
}
if (random2[j] < 0.08){
data[j,2] <- 1
}
}
s1 <- sum(data$gene == 1 & data$cancer == 1) # has the mutated gene & has cancer
s2 <- sum(data$gene == 1 & data$cancer == 0)
s3 <- sum(data$gene == 0 & data$cancer == 1)
s4 <- sum(data$gene == 0 & data$cancer == 0)
result[i,]$odd_withMutate <- s1/s2
result[i,]$odd_noMutate <- s3/s4
}
Different operations:
Operation #1:
If I run the code above, the 12th row of odd_noMutate in result will be 0, the 16th will be NaN. Then I tried to see what happened, so I use set.seed(12) or set.seed(16) to check (Operation #2 & #3). But the 0 and NaN disappeared! I mean, in Operation #2, 0.1638418 0 is not 1.5075377 0. In Operation #3, 0.2830189 0 is not 2.4013605 NaN.
Operation #2:
the changed part of code is:
set.seed(12) #odd_noMutate = 0
# set.seed(16) #odd_noMutate = NaN
# set.seed(i)
random1 <- runif(500, min = 0, max = 1)
set.seed(12)
# set.seed(16)
# set.seed(i) # add this instruction
random2 <- runif(500, min = 0, max = 1)
Operation #3:
# set.seed(12) #odd_noMutate = 0
set.seed(16) #odd_noMutate = NaN
# set.seed(i)
random1 <- runif(500, min = 0, max = 1)
# set.seed(12)
set.seed(16)
# set.seed(i) # add this instruction
Operation #4:
I find that even changed the i in my code will make results totally different (Shouldn't it be the subset of the original result?). It is the Operation #4. Specifically, 0.3092105 0 is not 1.5075377 0; 0.7562724 0 is not 2.4013605 NaN.
for (i in 10:20) {
# set.seed(12) #odd_noMutate = 0
# set.seed(16) #odd_noMutate = NaN
set.seed(i)
random1 <- runif(500, min = 0, max = 1)
# set.seed(12)
# set.seed(16)
set.seed(i) # add this instruction
random2 <- runif(500, min = 0, max = 1)
The results among these operations are shown below:

Answer 1

The problem is that some previous values in data remain and are reused. Maybe your problem are solved remaking data every for loop (i). (putting data <- data.frame(gene, cancer) into for loop).
gene <- vector(length = 500, mode = "integer")
cancer <- vector(length = 500, mode = "integer")
# data <- data.frame(gene, cancer)
odd_withMutate <- vector(length = 20, mode = "numeric")
odd_noMutate <- vector(length = 20, mode = "numeric")
result <- data.frame(odd_withMutate, odd_noMutate)
for (i in 1:20) {
data <- data.frame(gene, cancer) # remaking data every time
# set.seed(12)
# set.seed(16)
set.seed(i)
random1 <- runif(500, min = 0, max = 1)
# set.seed(12)
# set.seed(16)
set.seed(i) # add this instruction
random2 <- runif(500, min = 0, max = 1)
for (j in 1:500) {
if (random1[j] < 0.39){
data[j,1] <- 1
}
if (random2[j] < 0.08){
data[j,2] <- 1
}
}
s1 <- sum(data$gene == 1 & data$cancer == 1) # has the mutated gene & has cancer
s2 <- sum(data$gene == 1 & data$cancer == 0)
s3 <- sum(data$gene == 0 & data$cancer == 1)
s4 <- sum(data$gene == 0 & data$cancer == 0)
result[i,]$odd_withMutate <- s1/s2
result[i,]$odd_noMutate <- s3/s4
}
[ADDITION]
for loop doesn't have own environment unlike function.
So handling in for loop directly affects Global env objects such as your data.
You partly overwrited data of Global env by if statement and it was referred in next loop.
Here is my simple example;
data <- data.frame(gene = vector(length = 5, mode = "integer"))
keep_of_process <- list()
for(i in 1:2) {
set.seed(i)
random_val <- runif(5, 0, 1)
for(j in 1:5) {
if(random_val[j] < 0.39) {
data[j, 1] <- 1
}
keep_of_process[[i]] <- data.frame(random = random_val,
gene = data$gene)
}
}
do.call("cbind", keep_of_process) # just to merge process to show
# left is i = 1 and right is i = 2
random gene random gene
1 0.2655087 1 0.1848823 1
2 0.3721239 1 0.7023740 1
3 0.5728534 0 0.5733263 0
4 0.9082078 0 0.1680519 1
5 0.2016819 1 0.9438393 1
Please see row 2. In i = 2, random is 0.7023740 but gene is 1 (previous result retains).
So to do what you want (from my understanding), you need to remake data (my answer) or completely overwrite data by if statement, such as
if(random_val[j] < 0.39) {
data[j, 1] <- 1
} else {
data[j, 1] <- 0
}