I would like to compute these to quantities
a12=sum_(i from 1 to m)sum_(j1<j2)(I(X[i]>Y[j1] and X[i]>Y[j2]))
a13=sum_(j from 1 to n)sum_(i1<i2)(I(X[i1]>Y[j] and X[i2]>Y[j]))
where I is the indicator function.
So I came up with this R code
a12=0; a13=0
for (l in 1:(length(Z1)-1)){
for (m in 1:(length(Z2)-1)){
a12<-a12+(Z1[l]<Z2[m])*(Z1[l+1]<Z2[m])*1
a13<-a13+(Z1[l]<Z2[m])*(Z1[l]<Z2[m+1])*1
} # closing m
} # closing l
a12=a12+sum((Z1[-length(Z1)]<Z2[length(Z2)])*(Z1[-1]<Z2[length(Z2)])*1)
a13=a13+sum((Z1[length(Z1)]<Z2[-length(Z2)])*(Z1[length(Z1)]<Z2[-1])*1)
a12;
a13
Unfortunately, not only this is very slow but I am not getting what I am supposed to get.
Could you help me, please!
Thanks,
Roland
I'm assuming (for a12
) you want to do the following. You have two vectors x
(of length m
) and y
, and for each element x[i]
of x
, you are calculating the number of distinct index pairs j1
, j2
of y
such that x[i]
exceeds both y[j1]
and y[j2]
, and then you are summing this quantity over all i
. Here's a fast way to do a12
(the other will be left as an exercise). First note that you can flip the order of summation:
a12 = Sum_(j1 < j2) Sum_(i=1:m) I( X[i] > Y[j1] & X[i] > Y[j2] ),
ie for each distinct index-pair j1,j2
, we calculate the number of x
elements that exceed both y[j1]
and y[j2]
, and then we sum this quantity over all these distinct index-pairs. Now calculating the inner sum for pairs j1,j2
is like a matrix multiplication. Indeed, suppose we define vectors x
and y
:
set.seed(1)
x <- sample(1:5,5,T)
y <- sample(1:5,10,T)
then we can use the outer
function to produce a matrix y_x
whose [i,j]
entry is TRUE if and only if y[i] < x[j]
:
y_x <- outer(y,x,FUN = '<')
Now we get the inner sums by doing
z <- y_x %*% t(y_x)
where z[i,j]
is the number of elements of x
that exceed both y[i]
and y[j]
. Since we only want to sum z[i,j]
for distinct i < j
, we get the final result by taking the sum of the lower-triangle of z
using
a12 <- sum( z[lower.tri( z )])
> a12
[1] 72
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.