Data collection methodologies

In scientific studies, data collection can fall under two broad categories

  • observational study - where data is recorded without interfering with the process under study. These can be of 3 basic types
    • sample survey - provides info about a population at a particular point in time
    • prospective study - describes the population at present using sample survey and proceeds to follow forward in time to record specific outcomes
    • retrospective study - describes the population at present using sample survey and records specific occurrence that have already taken place
  • experimental study - where explanatory variables are actively manipulated and the effects on response variables are studied.

Sampling designs for surveys - observational study

The various ways in which data collection experiments can be designed:

  • Simple random sampling - Selecting a group of n units in such a way that each sample of size n has the equal probability of being selected.

  • Stratified random sampling - grouping the target population into strata based on a known auxiliary variable and then performing simple random sampling on each of the strata

  • cluster sampling - this involves selecting groups (based on convenience of study, such as buildings or city blocks) based on simple random sampling and then surveying or measuring all units within the groups.

  • systematic sampling - this involves selecting every nth value from a finite list as the sample. This technique is economical but can he heavily biased.

Design of experiments - experimental study

There are 2 types of variables in an experimental study - controlled variables (factors) and measured variables (response variables) which are under study

  • factorial treatment - each of the various factors (m,n,o..) affecting the phenomena are combined with one-another. The resulting combinations (mno..) are studied individually

  • fractional factorial treatment - as the number of factors increase, factorial treatment becomes impossible to complete. Then, only a few selected combinations are picked for the study forming a fractional factorial treatment.