Casual Research (experimentation)

Research

Exploratory

Descriptive

Causal

Experimentation

  1. Most interesting questions in marketing (and life) are causal Qs
    • Does higher price - more $?
    • Sun bathing cause skin cancer?
  2. Alot of confusion around causality
  3. Focus of today
    • procedures involved in causal research
    • why these procedures are necessary

process of manipulating the independent variable/ measuring its effect on the DV, while controlling for the effect of everything else

you need to offer X at different values/ obtain the corresponding values of Y
Option 1: Cre8 sub grps using your sample/apply different treaments (X) to diff grps
Option 2: Apply different treatments (X) over time

Experimental Group: grp exposed to the manipulated IV

Control Group: exposed to status quo value of the IV

Main obstacle: Extraneous Variables (EV)

  • variables, other than IV, influence the DV
  • Examples: diff in purchasing habits btwn customers, seasonality

Common sources of EV

  • poorly designed experiment stimuli
  • treatment/control grp X comparable
  • external events

Selection Bias

Random assignment of testing units -- eliminate selection bias

Commonly used Experimental Designs

One-Group Pretest-Posttest Design

Static Group Design

Posttest only Control Group Design

O1 X O2

  • grp of test unit measured twice
  • X control grp
  • treatment effect is computed as O2-O1
    -validity of this conclusion is questionable since EV are largely uncontrolled

EG: X O1
CG: O2

  • 2 group experimental design
  • EG exposed to treatment/ CG X
  • measurement on both grps made only after treatment
  • test unit X assigned at random
  • treatment effect O1-O2

EG: R X 01
CG: R X 02

  • similar to SGD, test units are randomly assigned to EG/CG
  • treatment effect is obtained by TE=O1-O2
  • selection bias is eliminated by randomisation

Classification of Exp Design

No RANDOMISATION

Pre-Experimental

  • 1 group Pretest-Posttest
  • Static Group

Quasi-Experimental

  • Time Series
  • Multiple Time Series

With RANDOMISATION

True Experimental

Posttest-Only Control Group

Statistical

Factorial Design

Repeated measurement
X random assignment

  • complex design test multiple IVs/ interactive effects
  • involve random assignment

Lab vs Field Exp

Lab Experiment

  • Takes place in Lab
  • Hypothetical scenarios
  • Research Participation

Field Experiment

  • Takes place in natural environment
  • Involves "real" decisions
  • A/B testing

Exp Design any good?

Internal Validity

External Validity

  • manipulation of IV is real cause of obs effect
  • degree to which effects of extraneous variable are controlled
  • can effects found be generalised in reality?
  • often comes down to lab vs field exp
  • do target population/ test units match?

Factors that -- poor internal validity= different EV

  • poorly designed experimental stimuli
  • treatment/ control grps X randomly assigned
  • treatment/ control take place at diff time (external events)

Limitations of Exp

Time Consuming

Expensive

Difficult to adminster

Competitors may deliberately contaminate the results of field exp