Sampling.md

Sampling

Let's create our own sample space using the Rev language. We'll begin by creating a vector in RevBayes. There are a few different ways to create a vector, but the simplest is probably to enclose a comma-separated list of values in square brackets [ . . . ]

sampleSpace = ["a","b","c"]

As with Python, to check the value of any variable, we can simply type the name of that variable and RevBayes will print its current value.

sampleSpace

In Rev, we can verify the size of our vector (i.e., the number of values it contains) by using its size() method.

sampleSpace.size()

Also like Python, we can check individual values in a vector by using the name of the vector and providing indices. For instance, to look at only the 2nd value in the list we would type

sampleSpace[2]

NOTE that RevBayes (like R) is 1-indexed, while Python is 0-indexed.

To sample values from our sample space, we need to know a few more things about the Rev language. First, if we want to create a new vector, we can do so implicitly by starting to assign values to individuals positions in the new vector. For instance, if we wanted to store a vector of values (called mySamples) sampled from our sample space, we could start by doing something like this

mySamples[1] = "a"
mySamples

Rev also contains for loops. To indicate a loop over a series of consecutive integers, it uses a convenient notation - the start and end values are just separated with :. Code executed as part of the loop is contained inside curly braces { . . .}.

for (i in 1:4){
    i
}

Now let's set up a for loop to do something more interesting, but we need one more trick. We're going to use a function called runifInt(). This name can be broken up into these three parts

1. r (sample values from a distribution)
2. unif (in this case, use a uniform distribution)
3. Int (more specifically, use a uniform distribution across only integers)

A uniform distribution just means that we assign equal probability to all values included in the distribution. runifInt() takes three arguments

1. the number of samples you want
2. the lower bound on your uniform distribution
3. the upper bound on your uniform distribution

Using a for loop, we can see what it looks like to sample from this uniform distribution several times.

for (i in 1:10){
    runifInt(1,1,5)
}

While the specific output will vary each time you run this loop, because the sampling process is stochastic, you should see a series of values sampled roughly evenly between 1 and 5. One last thing to note, though, is that even when you only want to sample a single value from this distribution, the result is returned as a list. A return value is the thing that a function hands back to you when you call it. Sometimes, we just want to pull that single number out by itself, so that it's not nested in a list. Because we know runifInt() will return a vector, we can add a vector index to pull out the single number.

for (i in 1:10){
    runifInt(1,1,5)[1]
}

Do you see how the output changes? Ok, now it's time to put this all together. We are going to create a for loop that will draw samples from our sample space and create a new vector called mySamples to store the sampled values. Can you use each of the pieces of syntax we talked about above to break down how this for loop works?

for (i in 1:4){
    mySamples[i] = sampleSpace[runifInt(1,1,sampleSpace.size())[1]]
}
mySamples

One additional thing to bear in mind as you run RevBayes code is that it's storing all of your variables in memory as you execute commands. Sometimes you want to wipe everything away to start fresh. To do that, you can use the clear() function. If you run this, your Rev slate will be clean.

clear()
mySamples