CODEWITHSUNDEEP
rdistribution
user20886
user20886

Reputation:

creating a function to calculate auc in R

I am extremely new with R, I have an assignment that I'm working on that I am having a lot of trouble with. I have defined a discrete probability distribution: 

s   P(s)
0   1/9
1   4/9
2   1/9
3   0/9
4   1/9
5   0/9
6   0/9
7   1/9
8   0/9
9   1/9

Now I have to work on this question:

Consistent with other distributions available in R, create a family of support functions for your probability distributuon:

f  =  dsidp(d)      # pmf - the height of the curve/bar for digit d
p  =  psidp(d)      # cdf - the probability of a value being d or less
d  =  qsidp(p)      # icdf - the digit corresponding to the given 
                    # cumulative probability p
d[]  =  rsidp(n)    # generate n random digits based on your probability distribution.

If someone could help me get started on writing these functions, it would be greatly appreciated!

Upvotes: 0

Views: 456

Answers (1)

Sven Hohenstein
Sven Hohenstein

Reputation: 81693

Firstly, read the data:

dat <- read.table(text = "s   P(s)
0   1/9
1   4/9
2   1/9
3   0/9
4   1/9
5   0/9
6   0/9
7   1/9
8   0/9
9   1/9", header = TRUE, stringsAsFactors = FALSE)

names(dat) <- c("s", "P")

Transform the fractions (represented as strings) to numeric values:

dat$P <- sapply(strsplit(dat$P, "/"), function(x) as.numeric(x[1]) / as.numeric(x[2]))

The functions:

# pmf - the height of the curve/bar for digit d
dsidp <- function(d) {
  with(dat, P[s == d])
}

# cdf - the probability of a value being d or less
psidp <- function(d) {
  with(dat, cumsum(P)[s == d])
}    

# icdf - the digit corresponding to the given cumulative probability p
qsidp <- function(p)  {
  with(dat, s[sapply(cumsum(P), all.equal, p) == "TRUE"][1])
}

Note. Since some probabilities are zero, some digits have identical cumulative probabilities. In these cases the lowest digit is returned by function qsidp.

# generate n random digits based on your probability distribution.
rsidp <- function(n) {
  with(dat, sample(s, n, TRUE, P))
}

Upvotes: 1

Related Questions