Initial values extraction for Odd Weibull distribution
Initial values extraction for Odd Weibull distribution
This function can be used to extract initial values found with empirical time on test transform (TTT) through initValuesOW function. This is used for parameter estimation in OW distribution.
param.startOW(param, initValOW)
Arguments
param: a character used to specify the parameter required. It can take the values "sigma" or "nu".
initValOW: an initValOW object generated with initValuesOW
function.
Returns
A length-one vector numeric value corresponding to the initial value of the parameter specified in param extracted from a initValuesOW
object specified in the initValOW input argument.
Details
This function just gets initial values computed with initValuesOW
for OW family. It must be called in sigma.start and nu.start
arguments from gamlss function. This function is useful only if user want to set start values automatically with TTT plot. See example for an illustration.
Examples
# Random data generation (OW distributed)y <- rOW(n=500, mu=0.05, sigma=0.6, nu=2)# Initial values with TTT plotiv <- initValuesOW(formula = y ~1)summary(iv)# This data is from unimodal hazard# See TTT estimate from sampleplot(iv, legend_options=list(pos=1.03))# See the true hazardcurve(hOW(x, mu=0.05, sigma=0.6, nu=2), to=100, lwd=3, ylab="h(x)")# Finally, we fit the modelrequire(gamlss)con.out <- gamlss.control(n.cyc =300, trace =FALSE)con.in<- glim.control(cyc =300)(sigma.start <- param.startOW("sigma", iv))(nu.start <- param.startOW("nu", iv))mod <- gamlss(y~1, sigma.fo=~1, nu.fo=~1, control=con.out, i.control=con.in, family=myOW_region(OW(sigma.link="identity", nu.link="identity"), valid.values="auto", iv), sigma.start=sigma.start, nu.start=nu.start)# Estimates are close to actual values(mu <- exp(coef(mod, what ="mu")))(sigma <- coef(mod, what ="sigma"))(nu <- coef(mod, what ="nu"))