# Probability

## Slides

## Notes

Assume we present Mr. Espressohead with 8 cups of coffee (lattes, actually). We tell him that 4 are made with Illy beans and the other 4 are made with Lavazza beans. His task is to taste all 8 and then divide the 8 cups into 4 Illy and 4 Lavazza.

The null hypothesis is that he is guessing. Our task then is to determine what is the probability of a given outcome, given the null hypothesis \(H_{0}\).

Let's compute the probability that he could pick the correct 4 Illy cups (and by thus the correct 4 Lavazza cups) just by guessing.

The probability of a correct outcome is defined by the ratio \(N_{c}/N_{t}\) where \(N_{c}\) is the number of correct ways of dividing up the cups, and \(N_{t}\) is the total number of ways (whether they are correct or not) of dividing up the cups.

Only one division of the cups is correct, so the numerator of the ratio \(N_{c}=1\).

The denominator is equal to the total number of ways of dividing 8 cups into two groups of 4. This is equivalent to the number of ways of choosing 4 Illy cups out of 8 (since then the other 4 will be determined as Lavazza cups).

In general, the number of ways of choosing \(k\) things out of a possible set of \(n\) (irrespective of the order) is given by the following equation, known as "n choose k":

\begin{equation} {n \choose k} = \frac{n!}{k!(n-k)!} \end{equation}where in general, the factorial operator \(x!\) = \((x)(x-1)(x-2) \dots (1)\).

In **R** we can use the `choose(n,k)`

function:

Ntot = choose(8,4) print(Ntot)

[1] 70

So the our denominator is 70, and the ratio \(N_{c}/N_{t}\) defining the probability of choosing the correct configuration just by guessing, equals 0.014.

```
guess_prob <- 1/Ntot
print(guess_prob)
```

[1] 0.01428571

Thus there is a 1.4 % chance that he could get them all correct, just by guessing.

Is that low enough? If you did observe him getting them all correct, would you believe he has the ability? Or would you still think he could be guessing?

An important lesson to learn in frequentist inferential statistics is
that we can **never** prove the alternate hypothesis (\(H_{1}\), that he
is not guessing, i.e. that he has the ability). All we can ever say is
that the probability of the null hypothesis \(H_{0}\) is very low. All
we can ever say is that it's very unlikely that he is guessing.

We could still be wrong however. Even if there were an experiment where the probability of \(H_{0}\) was exceedingly low (e.g. \(p < 0.001\)) or 1 in 1,000, there is still a non-zero chance that one could observe that outcome, under the null hypothesis. All we can say is that this is unlikely.