Conditional Joint Probabilities.
Abstract
Formerly, for the solution of the conditional probability of a single predictand, its equivalent normal deviate (END) was obtained, under the assumption of multivariate normality, by linear regression on the END's of the predictors. For the joint probability of two predictands, the approach is to find the two corresponding END's by the same method, but in addition, to find the conditional correlation coefficient between the predictands. Such correlation has proved to be the well-known partial correlation. In a few test examples, the conditional correlation has decreased significantly from the more basic unconditional correlation. However, the conditional correlation has remained large enough to make the conditional probabilities significantly higher than the mere product of the two marginal probabilities. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 02, 1978
- Accession Number
- ADA063817
Entities
People
- Irving I. Gringorten
Organizations
- Air Force Research Laboratory