Stručni rad

The Case of the Counterfeit Correlation: A Computer Simulation Illustrating a Common Error with the Use of Index Variables Including Per Capita Census Measures

Keith R. Billingsley ; The University of Georgia, USA
Robert Munzenrider ; The University of Florida, USA
James E. Prather ; Georgia State University, USA
Marwin K. Hoffman ; Appalachian State University, USA

Sažetak

The presented material is particularly directed to scholars who examine aggregate data i.e. index variables through the use of correlation techniques. As K. Pearson already demonstrated a substantial portion of the correlation between the underlying concepts but rather to the operation — division — by which the indices were created. He termed this correlation as »spurious« and demonstrated that it is due in part to the prior correlation of the variables forming the indices. Here we present a brief overview of the mathematics supporting the assertion of spuriousness and also the results of a computer simulation which we developed to demonstrate the dangers of the approach. Our simulations, hopefully, demonstrate that counterfeit correlations should be of practical concern to the researcher interested in actual applications and that correlating index variables is likely to lead to misleading conclusions because such correlations are counterfeit.

Ključne riječi

Hrčak ID:

156477

URI

https://hrcak.srce.hr/156477

Datum izdavanja:

31.12.1973.

Podaci na drugim jezicima: hrvatski

Posjeta: 1.209 *