Bob Luttman, Robert Luttman & Associates
Sample Size
Operational Issues: Documentation and Reporting
Traditional Pathway / Variance Documentation
Sample Size
If you work in a large urban teaching hospital that does hundreds of cardiac surgeries, deliveries, and joint transplants every year you don't have many sample size issues. However, if you work in a smaller hospital or a lower volume service, your variance analysis is limited by the sample size.
You simply cannot have more variables than you have patients.
The Statistics Gods outlawed that years ago.
This actually affects the analysis in two ways:
- Statistical models divide the degrees of freedom of a sample (usually the sample size minus 1) across the variables in the analysis and an estimate of the variation in the outcome measure. You cannot have more variables or do a deeper analysis than you have degrees of freedom. Even if you have enough samples to cover all the variables you could lose the ability to study interaction effects (is Variance A the problem? Variance B? or Both?).
- The smaller the sample size the wider the confidence interval, and therefore the less precise your results.
And the less likely that this year was better than last year. That leads us to the Halo Effect on the next page.
