R/tdForm.R

In BIOdry: Multilevel Modeling of Dendroclimatical Fluctuations

tdForm <- structure(function#ltd formulas
### This function formulates linear time-decline formulas (ltd) from
### categorical variables in multilevel ecological data series.
                    ##details<< the ltd formulas belong to following
                    ##general equation: log (x) = log (csx) + f(time);
                    ##where the relative organic growth (x) is
                    ##explained by the cumulative organic growth (csx)
                    ##plus a function of time f(time); with f(time)
                    ##being either the time or a logarithmic
                    ##transformation the time. The ltd can be
                    ##implemented by \code{\link{modelFrame}} function
                    ##to subtract trends in organic MEDS

                    ##references<< Zeide B. 1993. Analysis of Growth
                    ##Equations. For. Sci., 39: 594-616.
(
    rd, ##<<\code{data.frame} or \code{character} vector. Multilevel
        ##ecological data series or vector of ecological factors.
    prim.cov = FALSE, ##<<\code{logical}. Print a primary covariate
                      ##form: \code{'~ cov'}. If FALSE then a complete
                      ##formula: \code{'resp ~ cov | group'} is printed.
    on.time = TRUE, ##<< \code{logical}. If TRUE then \code{t =
                    ##'time'} (see \code{\link{rtimes}}). If FALSE
                    ##then \code{t = 'year'}.
    log.t = FALSE, ##<< \code{logical}. If TRUE then \code{f(time) =
                   ##ln(time)}. Default FALSE produces a log-linear
                   ##time-decline formula.
    lev.rm = NULL ##<< NULL or \code{character} name of the ecological
                  ##factor(s) in the MEDS to be removed from the
                  ##formula.
) {   
    rs <- 'log(x)'; lx <- '~ log(csx) +'; 
    t <- 'time'
    if(!on.time)t <- 'year'
    if(log.t)t <- paste('log(',t,')',sep = '')
    ftt <- paste(rs,lx,t,sep = ' ')
    if(!is.character(rd)){
        f <- cClass(rd, 'factor')
        if(is.numeric(lev.rm))
            lev.rm <- f[lev.rm]
        nf <- rev(f[!f%in%lev.rm])}
    if(is.character(rd))
        nf <- rd[!rd%in%lev.rm]
    sep. <- ' | '
    if(length(nf) == 0)
        sep. <- ''
    fc <- paste(nf,collapse = '/')
    fr <- paste(ftt,fc,sep = sep.)
    if(prim.cov)
        fr <- paste(lx,t,sep = '')
    return(formula (fr))
### \code{formula} with the forms: 'resp ~ cov | group' or '~ cov'.
} , ex=function(){
    ## an ltd formula:
    lev <- c('plot','tree')
    tdeq <- tdForm(lev,log.t = TRUE)
    tdeq
    ## (not run) only primary covariate:
    tdeq1 <- tdForm(lev,prim.cov = TRUE)
    tdeq1
    ##Multilevel data frame of tree-ring widths:
    data(Prings05,envir = environment())
    ## removing two levels: 'plot' and 'tree' from the formula
    tdea2 <- tdForm(Prings05, lev.rm = c('plot','tree'))
    tdea2 <- tdForm(Prings05, lev.rm = 2:3)
})