Experimental Statistics

From SkepticWiki

[edit] Definition

Experimental Statistics usually refers to that branch of mathematics applied to the analysis of data from experiments. An experimenter manipulates an independent variable to see if it has any reliable effect on a dependent variable.

[edit] Example

Either a drug or a placebo (independent variable) is randomly given to 14 people and then they are checked to see how many have died from Asia Bird Flu '05 a year later (dependent variable). Out of the 7 placebo subjects, all survived, but 5 of the drug group died.

A 2 X 2 Table where D = drug, P = placebo, A = alive and D = ex parrot:

	D	P
A	2	7
D	5	0

Then a link to Fisher's Exact Probability test would help.

This is called inferential statistics, because it allows one to evaluate the probability of an experimental or other result occurring by chance.

[edit] Discussion

These statistics mean nothing if the experiment is flawed and contains confounding variables correlated with the independent variable. GIGO