Statistics - (Scientific) Control (Group)

1 - About

Controls would show you things that have happened that should have not happened or things that didn't happened that should have happened. The first is called negative controls and the second is called positive controls

3 - Groups

The experimentation group and control group must be defined through careful experimental design:

  • randomized, (The creation of the group is randomized)
  • blinded, double-blinded (The subjects don't know on which group they belong)

3.1 - Random Sample

Experimental research requires that the creation of the groups (samples) are randomized.

This randomization is made through two steps:

  1. Random selection: A sample is created with individuals that are randomly selected from the population
  2. Random assignment: The individuals of the sample are randomly assigned to (conditions|groups)

The major benefit of randomized experiments is that they allow to make:

  • a strong claims about causality
  • an (inference|generalize) about that population

If it's truly random, we should get an uniform distribution for each groups:

3.2 - Double blind

A Double blind experiment occurs when the subject and the experimenter don't know to which group the subject belongs.

4 - Documentation

data_mining/control.txt · Last modified: 2014/09/21 17:42 by gerardnico