Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Sequential analyses in psychological research using Bayesian statistics

Klintefors, Pierre LU (2019) PSYP02 20191
Department of Psychology
Abstract
It is important in psychological research to use well planned methods that are as time and resource efficient as possible, without jeopardizing the reliability and validity of psychological science. The present paper aims to test how sequential analyses could be implemented in psychological research using Bayesian statistics. With sequential analyses it is possible to stop an experiment or study in the data collection stage for success or futility. To avoid offset estimation and false alarms, a mixture of model testing with Bayes Factor and Bayesian parameter estimation were used as stopping rules. After several runs of Monte Carlo simulations, it appears as a Bayes’ Factor (BF) boundary of 6 together with 95% Highest density interval... (More)
It is important in psychological research to use well planned methods that are as time and resource efficient as possible, without jeopardizing the reliability and validity of psychological science. The present paper aims to test how sequential analyses could be implemented in psychological research using Bayesian statistics. With sequential analyses it is possible to stop an experiment or study in the data collection stage for success or futility. To avoid offset estimation and false alarms, a mixture of model testing with Bayes Factor and Bayesian parameter estimation were used as stopping rules. After several runs of Monte Carlo simulations, it appears as a Bayes’ Factor (BF) boundary of 6 together with 95% Highest density interval (HDI) width under a SD*0.60 served as suitable stopping rules under conditions of simulations. However, the generalizability is limited by the simulations settings and the stopping rules are recommended to be implemented on data from real conducted experiments. (Less)
Please use this url to cite or link to this publication:
author
Klintefors, Pierre LU
supervisor
organization
course
PSYP02 20191
year
type
H1 - Master's Degree (One Year)
subject
keywords
Sequential analysis, Bayesian statistics, Optional stopping, Bayes' factor, Bayesian estimation, Highest Density Intervals
language
English
id
8993740
date added to LUP
2019-09-04 11:26:05
date last changed
2019-09-04 16:26:47
@misc{8993740,
  abstract     = {{It is important in psychological research to use well planned methods that are as time and resource efficient as possible, without jeopardizing the reliability and validity of psychological science. The present paper aims to test how sequential analyses could be implemented in psychological research using Bayesian statistics. With sequential analyses it is possible to stop an experiment or study in the data collection stage for success or futility. To avoid offset estimation and false alarms, a mixture of model testing with Bayes Factor and Bayesian parameter estimation were used as stopping rules. After several runs of Monte Carlo simulations, it appears as a Bayes’ Factor (BF) boundary of 6 together with 95% Highest density interval (HDI) width under a SD*0.60 served as suitable stopping rules under conditions of simulations. However, the generalizability is limited by the simulations settings and the stopping rules are recommended to be implemented on data from real conducted experiments.}},
  author       = {{Klintefors, Pierre}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Sequential analyses in psychological research using Bayesian statistics}},
  year         = {{2019}},
}