Does your stats gruntwork for you

Prior Belief

How likely this thing seemed before you started making observations. As the saying goes, "extraordinary claims require extraordinary evidence," and this measures how extraordinary the claim is. For example, if the hypothesis is that you have a specific disease, the prior is the frequency of that disease in the general population. If you can't measure directly, a good guideline is that the more complicated the hypothesis is, the lower this should be.

Conditional Likelihood

How likely this observation is assuming the hypothesis is true. Often, this can be rounded to 1.

Overall Likelihood

How likely this observation is in general, regardless of whether the hypothesis is true. If the hypothesis is "that man is scandanavian" and the observation is "he's blond", then this is the fraction of all men with blond hair.

False Positive Rate

How likely this observation is assuming the hypothesis is false. Sometimes the overall likelihood is hard to find, but we can calculate if for you using this and the other information given.

Normal Distribution

Sometimes the false positive rate is also hard to find. If the observation comes from a normally distributed population, we can help with that too. For example, if the observation is that a man is very short, you can enter the mean and standard deviation of male heights in his country, and his height, and we'll compute the likelihood of his being that short without special circumstances. Normal distributions are very common in nature, though not universal.

Binomial Distribution

If the observation was something like "we rolled this 4-sided die 10 times and 5 of them rolled 4s" we can calculate the false positive rate for that too. In that example, the values would be tries=10 normal_rate=0.25 successes=5.

More Observations

You can provide multiple observations that say things about your hypothesis. There are two caveats. First, the observations must be independant. For example, observing that two experts support something when in fact the second expert based his advise on reading the first one's statement is no good. Second, you must include all observations. If you make a lot of observations and only calculate based on the ones you like you'll conclude what you like, rather than what's true.