mind (version 1.1.0)

predict.mind: Predict method for Multivariate Linear Mixed Model

Description

Predicted values based on Multivariate Linear Mixed Model object

Usage

# S3 method for mind
predict(object,data,type= "proj",dir_s=NULL,dir_cov=NULL,...)

Value

When the predictor type is set equal to proj or synth the predict.mind produces a data.frame of predictions with the columns name equal to the multivariate responses. If the predictor type is eblup a data frame with predictions for all the profile (cross-clafficiation of domain of interest and coavariate patterns) is provided.

Arguments

object

an object of class "mind"

data

an object in which to look with which to predict.

type

the type of prediction required. ; the predictors type are the "eblup", "proj" and "synth".The default to "proj".

dir_s

optionally, if type is equal to eblup a data frame with the count for each modalities of the responce variable for the covariate patterns must be provided.

dir_cov

optionally, if type is equal to eblup a data frame with the sample units count for the covariate patterns must be provided.

...

arguments based from or to other methods

Author

Developed by Andrea Fasulo

Details

predict.mind produces predicted values, obtained by means the regression parameters on the frame data. In the actual version of predict.mind only unit level predictions are provided. If the type is equal to proj or synth a data.frame with individual predictions for all the modalities of the responce variable will be produced. When the eblup predictor is chosen two more input data.frame must be provided and the function will produce predictions for all the profiles identified by the cross-classification of the domains and covariate patterns.

Examples

# Load example data
data(data_s);data(univ)

# The sample units cover 104 over 333 domains in the population data frame
length(unique(data_s$dom));length(unique(univ$dom))
# \donttest{

# One random effect at domain level
formula<-as.formula(cbind(emp,unemp,inact)~(1|mun)+
factor(sexage)+factor(edu)+factor(fore))

# Drop from the universe data frame variables not referenced in the formula or in the broadarea
univ_1<-univ[,-6]

example.1<-mind.unit(formula=formula,dom="dom",data=data_s,universe=univ_1)

## Example 1
#Projection predictions
example.1.predict<-predict.mind(object=example.1,data=univ,type= "proj")

# Check if the sum of the unit level predictions at 
# domain level are equal to the mind.unit Projection predictions
ck<-cbind(univ,example.1.predict)
ck<-aggregate(cbind(emp,unemp,inact)~dom,ck,sum)
head(ck);head(example.1$PROJ)

## Example 2
#Synthetic predictions
example.1.synth<-predict.mind(object=example.1,data=univ,type="synth")
  
# Check if the sum of the unit level predictions at 
# domain level are equal to the mind.unit Synthetic predictions
ck<-cbind(univ,example.1.synth)
ck<-aggregate(cbind(emp,unemp,inact)~dom,ck,sum)
head(ck);head(example.1$SYNTH)

## Example 3
#EBLUP predictions
inp_1<-aggregate(cbind(emp,unemp,inact)~dom+mun+sexage + edu + fore,data_s,sum)
  
inp_2<-aggregate(emp+unemp+inact~dom+mun+sexage+edu +fore,data_s,sum)
  
example.1.eblup<-predict.mind(object=example.1,data=univ_1,type="eblup",dir_s=inp_1,dir_cov=inp_2)
  
# Check if the sum of the predictions at 
# profile level are equal to the mind.unit Eblup predictions
ck<-aggregate(cbind(emp,unemp,inact)~dom,example.1.eblup,sum)
head(ck);head(example.1$EBLUP)

# }