Next: rmvnorm() Up: Multivariate Macros Help File Previous: mvngen()   Contents


probsquad(x,structure(Q,L,addcon,grandmean) [, prior:P]), REAL matrix x
  with no MISSING elements, argument 2 a structure as returned by macro
  discrimquad(), REAL vector P of prior probabilities with P[i] > 0,
  default = rep(1/g,g)

Keywords: classification, discrimination
Suppose info = structure(Q,L,addcon,grandmean) as computed by
discrimquad(groups, y) summarizes quadratic discriminant functions
estimated from a n by p matrix of data from g groups.

probsquad(x, info), where x a REAL vector of length p, returns the
length g vector of posterior probabilities, assuming multivariate
normality of each groups and equal prior probability 1/g for each group.

probsquad(x, info, prior:P), where P is a length g vector with P[j] >
0, does the same using P[j]/sum(P) as prior probability of group j.

In both these usages x can be a m by p matrix, each row of which is an
observation and probsquad() returns a m by g matrix of posterior

See probsquad() for information about the form of info.

Gary Oehlert 2003-01-15