If you're throwing a dice on a table, then the entire universe could be the system, and then it generates the outcome say when you roll it.
A system should be thought as a finite list of rules that when followed results in the creation of something physically observable
Conditional indpencence, , suppose that we have a coin c1 with 0.5 for heads, and c2 with 0.9 for heads, so if we have these two coins, then if we know we're flipping coin c1, then we know the probability is 0.5. So we have that and .
If a person randomly picks a coin (and uses it for the rest of the time) an tosses it 6 times, can you guess the picked coni by observing the tosses of the coin? Well if it's
Suppose that we have H T T H T H, as a result of our tosses. Let toss denote this sequence, note that there is a
We found that and that . Note that we can generalize this idead for coins.
We had two models for each coin, then we observed the coin toss, and we were able to have a probabilty of which model is used. Note that the denominator of are all the same, which means we can compute
If the probability of each is the same, then this divisionis one, if the value is less than 1, then is more likely model, otherwise the value is greater than one, and then we know . The above is the Frequentist approach
is the amount of data in table required, and is the number of things it depends on, which is good because we can have being less than