Designing training experiments Flashcards
THE AIM
In Experimental studies the goal is to make an
inference about the causal effect of the IV (e.g., level of physical activity) on the DV (e.g., health)
This causal link can be clouded by certain factors:
confounders and mediators
CONFOUNDERS
An alternative IV that can explain the effect you
observed
MEDIATORS
An IV that partially determines
the strength of the relationship between your IV and DV
How to deal with confounders and mediators
Hold constant in sample Choose participants so your groups are matched for potential confounders (e.g., age, gender, etc). Hold constant in analysis Include potential confounding variables in your analysis as covariates
classic experimental designs
pre-post single
pre-post multiple
The 5 design issues
Pre/post test Training Control group Transfer Retention
The Hawthorne Effect
A phenomenon where behaviour is altered due to
the knowledge that one is being observed or monitored
It is a motivational effect, and therefore often leads
to improvements in performance (temporary)
Hawthorne effect
Wolfe and Michaud (2010)
264 rheumatoid arthritis patients
received comprehensive care in a clinical trial, including free treatment, and received ordinary RA care by their non-study physicians after the trial
Measured four variables pre-trial, post-trial, and eight months after post-trial (retention)
Almost half of the improvement noted in the clinical
trial HAQ score disappeared on entry to a non-sponsored follow-up study (8 months after post-trial), and from 23% to 44% of improvements in pain, patient global, and fatigue also disappeared
The Placebo Effect
The belief that you are receiving something to help performance can often increase motivation, confidence, etc.
Particularly problematic if your control group receives no training
The Placebo effect
Guillot et al (2012)
Examined the effect of motor imagery (MI) on tennis
serve performance
3 training groups:
Control – physical practice only
MI – physical practice plus motor imagery
Placebo – same as MI except were also told that a
custom-made racquet was developed for them
Both MI and placebo had significantly higher % of
successful serves than the control group
Only the placebo group had significantly more
accurate serves than the control group
Regression to the mean
Training programs often focus on players who are performing below their normal ability
There is a natural tendency for a player to move towards his/her average (mean) performance regardless of training method
Regression to the mean
Does sacking the manager improve football team’s performance? Anderson and Sally (2013)
“Even without sacking the manager, the performance of the control group bounces back in the same fashion and at least as strongly as the performance of the clubs that fired their managers (experimental group)”
The replication problem
findings in psychology have been called into question when follow-up studies failed to replicate them
May be due to:
False positive in original
False negative in replication
Correct but different results in both due to unintended methodological differences
The replication problem study by open science collaboration (2015)
100 experimental and correlational studies published in psychology were replicated
36% of replicated studies were significant, compared to 97% in original studies
Mean effect size was less than half those of the original results
What are pre-mortems?
Pre-mortems occur before a project commences
and allow individuals to identify what might cause the project to “die”
Shown to be effective in combating overconfidence
in project planning above other techniques (e.g. ‘Pros/Cons Generation’)