Return EM algorithm output for right, left or interval censored regression model under SMN distributions (Normal, Student-t, Pearson VII, Slash or Contaminated Normal), built the corresponding envelope graph and compute some criteria for model selection, such as AIC, BIC and EDC.
CensReg.SMN(cc,x,y,LS=NULL,nu=3,delta=NULL,cens="left",dist="T",
show.envelope="FALSE", error=0.0001,iter.max=300)Vector of censoring indicators. For each observation: 0 if non-censored, 1 if censored.
Matrix or vector of covariates.
Vector of responses in case of right/left censoring. Vector of lower limits if censoring is intervalar.
Vector of upper limits if interval censoring. Must not be provided in case of left/right censoring.
Initial value of the parameter of the scale variable of the SMN family. Must not be provided in case of Normal distribution. Must be a bidimensional vector in case of contaminated normal distribution (NormalC).
Second parameter of Pearson VII, fixed. Must not be provided in case of Normal, Student-t or slash distribution.
"left" for left censoring, "right" for right censoring, "interval" for interval censoring.
Distribution to be used in fitting: "Normal" for Normal model, "T" for Student-t model, "PearsonVII" for Pearson VII model, "Slash" for slash model and "NormalC" for contaminated Normal model.
TRUE or FALSE. Indicates if envelope graph should be built for the fitted model (based on transformed Martingale residuals). Default is FALSE.
The convergence maximum error.
The maximum number of iterations of the EM algorithm. Default=300.
EM estimates for the regression coefficients.
EM estimates for the scale parameters.
Returned the value of the log-likelihood under the fitted model.
Number of interations until convergence.
AIC criteria for model selection.
BIC criteria for model selection.
EDC criteria for model selection.
Standard error estimates.
For the contaminated Normal distribution, each component of the bidimensional vector "nu" must lie on (0,1). For the Pearson VII distribution, delta is fixed as the provided value and is not estimated. The parameters beta and sigma2 are initialized with the minimum square estimators of the regression x vs y. If you want to fit a regression model for non-censored data, just set "cc" as a vector of zeros and "cens" as either "right" or "left".
Aldo M. Garay, Victor H. Lachos, Heleno Bolfarine, Celso R. Cabral. "Linear Censored Regression Models with Scale Mixture of Normal Distributions". Statistical Papers.(2017) 58:247--278.
# NOT RUN {
##see examples in \code{\link{wage.rates}}
# }
Run the code above in your browser using DataLab