The judgment of covariation between binary variables: Some conditions that influence the process

Abstract

Two hundred subjects made decisions based on the data of two binary variables, ethnic origin and employment status. They were asked to specify the information necessary and sufficient for making the decision before doing so. Variables studied were: (1) Type of question. Five questions were used, four of which focused on one of the four instances involved (e.g., Which information is necessary to find out ifthere is a connection between Asian immigrants and unemployment?). Question five was unbiased. (2) Type of variable. Half of the subjects found two symmetric variables, and half found one symmetric and one asymmetric variable. (3) Level of correlation. Five different correlations were used, three favoring Asians, one favoring Europeans, and one representing a zero correlation. Subjects made logical decisions that were not influenced by any of the experimental conditions. The selection of information before and after making choices was influenced by all variables except level of correlation. The biased questions produced deviations from normative reasoning; the two symmetrical variables and the process of making decisions encouraged normative reasoning.

Self‐truncated sampling produces more moderate covariation judgment and related decision than descriptive frequency information: The role of regressive frequency estimation

“…It is worth noting that similar decision tasks had been used in Seggie and Endersby ( 1972 ) and Seggie ( 1987 ) to investigate whether people had a proper concept of correlation and the potential influential factors. However, their studies involved only a binary decision (i.e., whether to transport a local patient to a distant hospital or not) rather than a quantitative measure of RCP as in the current study.…”