Bioequivalence and Bioavailability Forum

Hi Sundar,

❝ […] to write programming in R or Maxima particularly, would proficiency in C, C++ or Java computer languages help.

Sure. Though some structures of R are derived from Lisp, the syntactical structure is very similar to the languages you mentioned. Furthermore, R allows linking to compiled subroutines in FOR­TRAN, C, C++ (see f.i. the RFortran project) if you want to speed things up.

The syntax of Maxima (core again derived from Lisp) is particular, but similar to other Computer Algebra Systems (Derive, Mathcad, …). Personally I use it sometimes when my algebra-skills are too limited – and implement the derived solution in R. One of the strengths of CASs is calculation with unlimited numeric precision – something you don’t really need in PK.

R will fail for any p >15 due to limited numeric precison (even in the 64bit version!):

p <- 15 # any >= 0
x <- rep(1*10^p, 3)+(-1:1)
# example: p = 1, x = (9, 10, 11)
# standard deviation should be /exactly/ 1
if(sd(x) == 1) {
  cat("pass:       SD(x) = 1\n")
  } else {
  cat("fail:       SD(x) != 1\ncalculated:", sd(x), "\n")
}
p <- 16
x <- rep(1*10^p, 3)+(-1:1)
if(sd(x) == 1) {
  cat("pass:       SD(x) = 1\n")
  } else {
  cat("fail:       SD(x) != 1\ncalculated:", sd(x), "\n")
}

Now in Maxima (p = 10,000‼):

load(descriptive)$
p: 10000;
x: ([1,1,1])*10^p+([-1,0,1]);
std1(x);
Which gives:
(%o2) 10000
(%o3) [899999999999999999999999999999[9941 digits]999999999999999999999999999999,
900000000000000000000000000000[9941 digits]000000000000000000000000000000,
900000000000000000000000000000[9941 digits]000000000000000000000000000001]
(%o4) 1

Amazing. I don’t know where Maxima’s limitations are, but 10^5,000,000 + {–1, 0, +1} gave still the correct standard deviation.