Distortion in the Pearson product moment correlation due to a restricted number of scale points is evaluated in two ways. First, a simulation of the bivariate normal distribution is used to estimate the effects of varying the number of scale points on the product moment correlation. This procedure demonstrates a substantial amount of information loss. Second, other correlation coefficients and some methods to correct for this loss are discussed and related to the simulation data.
Get full access to this article
View all access options for this article.
References
1.
FarleyJohn U., and HowardJ.Controlling “Error” in Data.Lexington: Lexington Press, 1975.
2.
GreenP. E., and RaoV.“Rating Scales and Information Recovery: How Many Scales and Response Categories to Use?”Journal of Marketing, 34(1970), 33–9.
3.
GuilfordJ. P.Psychometric Methods.New York: McGraw-Hill Book Company, 1954.
4.
LehmannD. R., and HubertJ.“Are Three Point Scales Always Good Enough?”Journal of Marketing Research, 9(November 1972), 444–6.
5.
MartinW. S.“The Effects of Scaling on the Correlation Coefficient: A Test of Validity,”Journal of Marketing Research, 10(August 1973), 316–8.
6.
McNemarQuinn. Psychological Statistics, 4th ed. New York: John Wiley and Sons, Inc., 1969.
7.
MorrisonDonald G., “Regression with Discrete Random Variables: The Effect on R2,”Journal of Marketing Research, 9(August 1972), 338–40.
8.
MorrisonDonald G.“Upper Bounds for Correlations Between Binary Outcomes and Probabilistic Predictions,”Journal of the American Statistical Association, 67(1972), 68–70.
9.
PearsonK.“On the Measurement of the Influence of ‘Broad Categories’ on Correlation,”Biometrika, 9(1913), 116–39.
10.
PearsonK., and PearsonE. S.“On Polychoric Coefficients of Correlation,”Biometrika, 14(1922), 127–56.
11.
PetersC. C., and Van VoorhisW. R.Statistical Procedures and Their Mathematical Bases.New York: McGraw-Hill Book Company, 1940.
