R/curvefit.R

In curvetest: The package will formally test two curves represented by discrete data sets to be statistically equal or not when the errors of the two curves were assumed either equal or not using the tube formula to calculate the tail probabilities.

curvefit <-
function(formula, data,  kernel="Quartic", alpha=0.5,  bw=NULL, 
      myx, bcorrect="simple", getit=F){    
     if(is.character(kernel)) {
      kernel.char=kernel
      kernel=get.weight.function(kernel)
     }else{kernel.char=attr(kernel, "name")}
    if(length(bcorrect)>1) bcorrect=bcorrect[1] 
    x<-model.frame(formula, data)[,2] 
    y<-model.frame(formula, data)[,1] 
    if(missing(myx)) myx=x else if(length(myx)==1) myx=seq(min(x), max(x), length=myx)
    if(!is.numeric(alpha)|is.null(alpha)) alpha=getoptimalalpha(formula, data)
    mf <- match.call(expand.dots = FALSE)
    m <- match(c("formula", "data",  "kernel", "alpha", "myx", "plotit"), names(mf), 0L)
    mf <- mf[c(1L, m)]
    mf[[1L]] <- as.name("model.frame")
    mf <- eval(mf[1:3], parent.frame())
    mt <- attr(mf, "terms")
    y <- model.response(mf, "numeric")
    x <- model.matrix(mt, mf)[,2]
    xlab<- deparse(formula[[3]])
    ylab<- deparse(formula[[2]])
    #
    o<-getWV(x=x, myx=myx, kernel=kernel.char, alpha=alpha, bw=bw, bc=bcorrect, getit=FALSE)
    x<-o$x;  myx<-o$myx; nn<-length(myx); n1<-length(x);  k0<-o$k0; 
    delta<-o$delta; delta2<-o$delta2;vv<-o$vv;TX<-o$TX; sx<-o$sx;
    LL<-o$LL; AA<-o$AA;
    ##############################0######    
    ey <-  LL%*% y
    bic<-log(sum(ey^2))*n1+log(n1)*delta/2 ##BIC statistic
    out<-list(formula=formula, data.origninal=data, data.model=data.frame(y=y, x=x), 
      xlab=xlab, ylab=ylab, fits=ey, 
      res=y-ey,  myx=myx, fitted=sx%*%y, BIC=bic,  k0=k0, vv=vv, 
      alpha=alpha, sx=sx,AA=AA, kernel=kernel, LL=LL, delta=delta,  delta2=delta2, bcorrect=bcorrect)
    class(out)<-"curvefit"    
    return(invisible(out))
 }