[R-sig-ME] Input function when using nlme() in nlme package

[Abdullah içen]

Hi all,

I've tried nlmer() function but it couldn't converge for some reason. Now
i'm trying to use nlme() function in nlme package my function is below

Sandikkaya_NLSite <- function(PSAr, Vs30, h800, b1, b2, b3, c){

  #Linear term
  .exprN1 <- ifelse(Vs30>1000, 1000/760, Vs30/760)
  .exprN2 <- log(.exprN1)
  .valueN <- b1*.exprN2

  # Depth-to-rock term (in log)   #!!!!!!!!!!!!!!!!!! z<1 = 1 i deleted it
dont forget to change!!!!!!!!!!!!!!
  # in SD18 they are estimated from CY08 equation for Z1
  h800[h800<1]<-1
  .exprZ1 <- ifelse(Vs30>=760, 1, h800)
  .exprZ2 <- log(.exprZ1)
  .valueZ <- b2*.exprZ2

  #Nonlinear term
  .exprNL1 <- ifelse(Vs30>760, 760-360, Vs30-360)
  .exprNL2 <- -0.00701*.exprNL1
  .exprNL3 <- exp(.exprNL2)
  .exprNL4 <- -0.00701*400
  .exprNL5 <- exp(.exprNL4)
  .exprNL6 <- .exprNL3 - .exprNL5
  .exprNL7 <- PSAr/c
  .exprNL8 <- 1+.exprNL7 #log(1+PSAr/c)
  .exprNL9 <- log(.exprNL8)
  .valueNL <- b3*.exprNL9*.exprNL6


  .value <- 0 + .valueN + .valueZ + .valueNL

  .grad <- array(0, c(length(.value), 4L), list(NULL, c("b1", "b2",
"b3","c")))

  ## e1 ##

  ## b1, b2 ,b3 ##
  .grad[, "b1"] <- .exprN2
  .grad[, "b2"] <- .exprZ2
  .grad[, "b3"] <- .exprNL9*.exprNL6
  .grad[, "c"] <- b3*.exprNL6*((1/(.exprNL8))*(-.exprNL7/c))


  attr(.value, "gradient") <- .grad
  .value
}



and i tried to run nlme() like

fm1 <- nlme(log(PSAm) ~ Sandikkaya_NLSite(PSAr, Vs30, h800, b1, b2, b3, c),
 data = data,
                       fixed = b1+ b2+ b3~1 ,
                       random = EQID~1 ,
                       start = c(b1 = 1, b2 = 1, b3 = 1, c=0.1))



b1,b2,b3 and c are my fixed effects and my random effects in lmer is
(1|EQID), EQID is earthquake ids in my data frame (data) my structure in
lmer is below


fm <-lmer(log(data$PSAm)~ 0  + data$VN + data$Z1 + data$VN03 + data$VN19 +
data$VN21 + data$VN28 + (1|data$EQID)  , data, REML=FALSE)


thank you for your attention.

	[[alternative HTML version deleted]]