Proportional odds and related models

Fits cumulative link models: proportional odds, probit, complementary log-log, and cauchit.

Usage

svyolr(formula, design, ...)
# S3 method for class 'survey.design2'
svyolr(formula, design, start, subset=NULL,...,
    na.action = na.omit,method = c("logistic", "probit", "cloglog", "cauchit"))
# S3 method for class 'svyrep.design'
svyolr(formula,design,subset=NULL,...,return.replicates=FALSE, 
    multicore=getOption("survey.multicore"))
# S3 method for class 'svyolr'
predict(object, newdata, type = c("class", "probs"), ...)

Arguments

formula

Formula: the response must be a factor with at least three levels

design

survey design object

subset

subset of the design to use; NULL for all of it

...

dots

start

Optional starting values for optimization

na.action

handling of missing values

multicore

Use multicore package to distribute computation of replicates across multiple processors?

method

Link function

return.replicates

return the individual replicate-weight estimates

object

object of class svyolr

newdata

new data for predictions

type

return vector of most likely class or matrix of probabilities

Value

An object of class svyolr

Author

The code is based closely on polr() from the MASS package of Venables and Ripley.

See also

svyglm, regTermTest

Examples

data(api)
dclus1<-svydesign(id=~dnum, weights=~pw, data=apiclus1, fpc=~fpc)
dclus1<-update(dclus1, mealcat=cut(meals,c(0,25,50,75,100)))

m<-svyolr(mealcat~avg.ed+mobility+stype, design=dclus1)
m
#> Call:
#> svyolr(mealcat ~ avg.ed + mobility + stype, design = dclus1)
#> 
#> Coefficients:
#>     avg.ed   mobility     stypeH     stypeM 
#> -2.6999217  0.0325042 -1.7574715 -0.6191463 
#> 
#> Intercepts:
#>   (0,25]|(25,50]  (25,50]|(50,75] (50,75]|(75,100] 
#>        -8.857919        -6.586464        -4.924938 

## Use regTermTest for testing multiple parameters
regTermTest(m, ~avg.ed+stype, method="LRT")
#> Working (Rao-Scott+F) LRT for avg.ed stype
#>  in svyolr(formula = mealcat ~ avg.ed + mobility + stype, design = dclus1)
#> Working 2logLR =  0.6580304 p= 0.73257 
#> (scale factors:  2.7 0.22 0.11 );  denominator df= 10

## predictions
summary(predict(m, newdata=apiclus2))
#>   (0,25]  (25,50]  (50,75] (75,100] 
#>       51       46       16       13 
summary(predict(m, newdata=apiclus2, type="probs"))
#>      (0,25]            (25,50]            (50,75]             (75,100]        
#>  Min.   :0.004597   Min.   :0.004944   Min.   :0.0004632   Min.   :0.0001086  
#>  1st Qu.:0.096088   1st Qu.:0.153546   1st Qu.:0.0353593   1st Qu.:0.0086690  
#>  Median :0.334762   Median :0.330429   Median :0.1326655   Median :0.0374619  
#>  Mean   :0.400508   Mean   :0.307402   Mean   :0.1658904   Mean   :0.1262000  
#>  3rd Qu.:0.691552   3rd Qu.:0.465418   3rd Qu.:0.2877986   3rd Qu.:0.1555944  
#>  Max.   :0.994485   Max.   :0.513732   Max.   :0.3927701   Max.   :0.8091764