Parrotfish in wrightscape

Bootstrapping option

Needs a method to pass the optional optim arguments to the update function, such that update uses the same algorithm as the initial fit. Required being a little clever:

[gist id=“1010999”]

Done. but perhaps we won’t always want to bootstrap under this? I suppose it is good practice but slow! Shoot, now I need a lot of really large computers really soon.

Excellent suggestion for better ways to do this from statscicomp list. match.call looks very promising as it avoids making a copy, but risks being evaluated in the wrong environment, which my solution above avoids.



# Using match.call()
#--------------------
f = function(y, ... ) {
 myCall = match.call()
 list(myCall = myCall, ans = y + dnorm(...))
<div>}
a = f(1, x = 4, sd = 2, mean = 4)> length(a$myCall)</div>
[1] 5
> a$myCall
f(y = 1, x = 4, sd = 2, mean = 4)
 > a$myCall[[1]] # the function
f
> a$myCall[[2]] # represents the first argument, y
[1] 1
# Now I want to change this second value (representing y) to a 2
> a$myCall[[2]] = 2
> a$myCall
 f(y = 2, x = 4, sd = 2, mean = 4)
eval(a$myCall)
$myCall
f(y = 2, x = 4, sd = 2, mean = 4)
$ans
[1] 2.199471
# Using closures
#---------------
set_up_f = function(...) {
 fFn = function(y) y + dnorm(...)
 }
f = set_up_f(x = 4, sd = 2, mean = 4)
f(1)
# 1.199471
f(2)
# 2.199471

Note: parallelization function can use pre-initialized parallelization, but should warn that it is using this setup and report how many processors are being used vs requested.

General against Brownie, General against Ouch: