A basic question on hypothesis testing In hypothesis testing a hypothesis is generally defined to be "a statement about the value of a population parameter". For example the mean value of the height of people living in a certain city.
I do not understand how this applies to the classical example of coin tossing. In coin tossing we test the hypothesis that the coin is fair. What is the population here and which parameter of this population is under concern?
Thanks
 A: Let's say you toss your coin 100 times. Then you count how often you've got heads. 
Doing this very often you may draw a graph that shows how often you've got one head, two heads, three heads ...up to 100 heads. Your x-axis is 1 to 100 heads, your y-axis is how often you've got it tossing very very often 100 times your coin.
With a perfect coin, you will find that for the most you have 50 times heads. This will be your parameter of central tendency. But of course sometimes there will be more or less than 50 heads, even using a perfect coin. Of course it very unlikely that you get only one head, when tossing 100 times. How (un)likely it is shows your graph.
When doing an empirical experiment with a specific coin, you may toss 100 times and count how often you really found head. Given the H0 that the coin is perfect, you can determine how likely is, what you found empirically. Just go to your graph and have a look.
The second parameter (spreading) depends on how often you toss your coin each trial. If your trial is not 100 tosses but let's say just 10 tosses, you will get more variance (and of course your mean value is 5 and not 50). So your H0-graph depends on how often you toss each trial.
However, the coins are not central but help to understand the idea. Doing a study you have a concrete number of observations ("tosses per trial") and you can calculate the parameters. You can determine how likely is what you found - given the H0.
A: The probability of seeing H heads from n coin tosses follows the Binomial distribution Bin(n,p) where p is the probability of seeing a head in a given coin toss. The Binomial test can be used to test the number of observed heads from n coin tosses against the expected number assuming the null hypothesis that the coin is fair. 
There are plenty of worked examples for such problems, for example one involving rolling a die at http://en.wikipedia.org/wiki/Binomial_test .
A: The population is all the tosses of the coin, the parameter is the frequency of the head/tails parameter. Notice that you won't be able to get a random sampling in these kind of situations and the population varies through time, so the experiment differs in important ways to the first one you presented. 
