How big sample do you need for norms?

Higher number give you better result, but can be slow.

© 2017 Hynek Cígler
Department of psychology, Faculty of Social Studies
Masaryk University, Brno, Czech Republic
Source code on GitHub

Sampling results:

It is well known the fact that if the sample drawn from a population is used for the population quantiles estimates, sampling error is higher in extreme quantilles of the distribution than in the mean scores. However, this sampling error distorsion is hard to imagine.

This simple tool was developed to simulate sampling error at difference level of distribution during the standardization of psychological tests. It uses the IQ scale for better understandability. The error includes only random sampling error; it does not account for non-representative samples.

The mean and standard deviation was used to quantiles estimates. The normal, non-skewed distribution was used.

  • Pirate plot: result of all the samples is drawn.
  • Cumulative: one sample is draw as the true and observed cumulative function.
  • Linear plot: one sample is draw as the relation between true and norm implied score.