Related Experiment Videos
How many subjects to screen? A practical procedure for estimating multivariate normal probabilities for correlated
K O McGraw1, S Gordji, S P Wong
1Department of Psychology, University of Mississippi 38677.
Journal of Consulting and Clinical Psychology
|October 1, 1994
Summary
This study presents a practical method for determining sample sizes needed to meet multiple criteria. It uses regression analysis to accurately estimate joint probabilities based on correlation, aiding in efficient sample selection.
Area of Science:
- Statistics
- Quantitative Psychology
- Research Methodology
Background:
- Determining appropriate sample sizes is crucial for studies with multiple correlated criteria.
- Existing methods may not adequately address the complexities of multivariate probability estimation.
- Accurate sample size estimation ensures the reliability and validity of research findings.
Purpose of the Study:
- To propose a practical procedure for estimating the required sample size for studies with multiple correlated selection criteria.
- To provide a method for predicting multivariate probabilities based on pairwise correlations.
- To enhance the efficiency of sample selection in research settings.
Main Methods:
- Utilizing least squares regression to model Monte Carlo estimates of multivariate probabilities.
- Plotting these estimates as a function of mean pairwise correlations (r) for criterion variables.
- Developing predictive equations for joint probabilities involving 3 to 5 variables.
Main Results:
- Least squares regression offers a good quadratic fit for Monte Carlo estimates of multivariate probabilities.
- The derived equations can predict selected 3- to 5-variable joint probabilities with reasonable accuracy.
- The procedure is effective when pairwise correlations for selection criteria range from 0.10 to 0.90.
Conclusions:
- The proposed procedure offers a practical and accurate approach to sample size estimation for complex criteria.
- This method facilitates the acquisition of samples that meet multiple, correlated requirements.
- Researchers can confidently apply these equations to optimize their sampling strategies.