Questions.md

Questions

• What is the class of NA? Why?

Consider the function

f =
function(x, len = length(x))
{
  x = x[!is.na(x)]
  sum(x)/len
}

Now,

a = rnorm(10)
a[c(3, 9)] = NA
f(a)

What is the value of n in sum(x)/n?

And in

f(a, 20)

• Why did I not use n as the name for the parameter len above?

• Consider the function definition

f = 
function(x, y)
{
   if(is.numeric(x) && all(x < 0))
     return(sum(x))
	 
    x ^ y
}

Now we call this with

 f(a, z <- 3)

What is the value of z at the end of the call?

You don't know a? Okay

a = rnorm(100, - 20, 4)

f(a, z <- 3)

What is z?

• We define a function as

toPDF =
function(imgFile, outFile = removeExtension(imgFile), 
         renderer = PDFRenderer(outFile, GetDataPath(api), ...),
         api = tesseract(, PSM_AUTO), ...)
{
   args = list(...)
   args$x
}

Can the default value of renderer be a call that references outFile? How will it find outFile? And how will it find the api variable?

• In the function toPDF() above, explain the three uses of ...

• Consider the function

area =
function(r)
{
   pi * r^2
}

What are the parameters? the local variables? the global/non-local variables? Where are the global variables found?

• Continuing from the previous question What is area(10)?

• Continuing, what is the result of

pi = 2
area(10)

• The gaussian() function is defined in stats package. It is used to define a gaussian family in a generalized linear model (see the help ?gaussian). It uses the variable pi. When we redefine pi as we did above, will gaussian() behave differently? Is this good or bad? If we wanted the opposite to occur, how could we do this?

f = 
function(x)
{
  function(theta)
     sum(log(dexp(x, theta)))
}

Where will f find theta?

• Continuing from the previous question

d = rexp(10)
lik = f(d)

What class of object is lik()?

• Continuing again

sapply(seq(1, 2, by = .01), lik)

What is theta in each call to lik().

• Some people think it would be a good idea if R found non-local variables by walking up the call stack. Why is this not a good idea?