R/FSAUtils.R
logbtcf.Rd
Constructs the correction-factor used when back-transforming log-transformed values according to Sprugel (1983). Sprugel's main formula -- exp((syx^2)/2) -- is used when syx is estimated for natural log transformed data. A correction for any base is obtained by multiplying the syx term by log_e(base) to give exp(((log_e(base)*syx)^2)/2). This more general formula is implemented here (if, of course, the base is exp(1) then the general formula reduces to the original specific formula).
logbtcf(obj, base = exp(1))
obj | An object from |
---|---|
base | A single numeric that indicates the base of the logarithm used. |
A numeric value that is the correction factor according to Sprugel (1983).
Sprugel, D.G. 1983. Correcting for bias in log-transformed allometric equations. Ecology 64:209-210.
# toy data df <- data.frame(y=rlnorm(10),x=rlnorm(10)) df$logey <- log(df$y) df$log10y <- log10(df$y) df$logex <- log(df$x) df$log10x <- log10(df$x) # model and predictions on loge scale lme <- lm(logey~logex,data=df) ( ploge <- predict(lme,data.frame(logex=log(10))) )#> 1 #> 1.691612( pe <- exp(ploge) )#> 1 #> 5.428224( cfe <- logbtcf(lme) )#> [1] 1.381993( cpe <- cfe*pe )#> 1 #> 7.501769# model and predictions on log10 scale lm10 <- lm(log10y~log10x,data=df) plog10 <- predict(lm10,data.frame(log10x=log10(10))) p10 <- 10^(plog10) ( cf10 <- logbtcf(lm10,10) )#> [1] 1.381993( cp10 <- cf10*p10 )#> 1 #> 7.501769# cfe and cf10, cpe and cp10 should be equal all.equal(cfe,cf10)#> [1] TRUEall.equal(cpe,cp10)#> [1] TRUE