The returned model evaluates to the fit of the priors in deviance (-2 log likelihood) units. The analytic gradient and Hessian are included for quick optimization using Newton-Raphson.
univariatePrior(type, labels, mode, strength = NULL, name = "univariatePrior")one of c("lnorm","beta","logit-norm")
a vector of parameters to which to apply the prior density
the mode of the prior density
a prior-specific strength (optional)
the name of the mxModel returned
an mxModel that evaluates to the prior density in deviance units
Priors of type 'beta' and 'logit-norm' are commonly used for the lower asymptote parameter of the 3PL model. Both of these priors assume that the parameter is in logit units. The 'lnorm' prior can be used for slope parameters.
# NOT RUN {
model <- univariatePrior("logit-norm", "x1", -1)
model$priorParam$values[1,1] <- -.6
model <- mxRun(model)
model$output$fit
model$output$gradient
model$output$hessian
# }
Run the code above in your browser using DataLab