dlmScala

Univariate DLMs in Scala

To Do

• discount factor

• Superposition (define + operator for DLM objects)

• filter method

• forecast method

• smooth method

• backSample method (see p. 114 of WH)

• Gibbs object (for FFBS)

• Merge Model class with DLM class

object DLM {
  abstract class Generic(F:Any,G:Any,V:Any,W:Any) {
    type Obs
    type ObsVar
    type Prior
    // Param is the posterior updates like in Param.scala
    type Param 
    // This is the type of the state parameter (vector, matrix)
    type State

    def filter(y:List[Obs],init:Param,prior:Prior):List[Param]
    def forecast(y:List[Obs],filt:List[Param],nAhead:Int=1):List[(Obs,ObsVar)]
    def smooth(y:List[Obs],filt:List[Param]):List[(Obs,ObsVar)]
    def backSample(y:List[Obs],filt:List[Param]):List[State]
  }

}

Interesting Excursions for Univariate and Multivariate?

• type unions?

type ~[A] = A=>Nothing
type V[U,T] = ~[ ~[U] with ~[T] ]
type ~~[A] = ~[~[A]]
type OR[T,U] = { type Lambda[X] = ~~[X] <:< (T V U) }

type OR[T,U] = {
  type ~[A] = A=>Nothing
  type V[U,T] = ~[ ~[U] with ~[T] ]
  type ~~[A] = ~[~[A]]
  type Lambda[X] = ~~[X] <:< (T V U)
}

def bla[T: (Int OR Double)#Lambda](x: T):Double = x match {
  case x:Int => x.toDouble
  case x:Double => x + 1
}

bla(10)
bla(11.1)

Or try this.