gamfit: GAM for environment-ordination fit

In phytomosaic/foggy: Species and Trait Responses to Fog Gradients

Description

Generalized Additive Model (GAM) to regress one or many environmental variables on NMDS ordination scores.

Usage

gamfit(ord, env, ...)

## S3 method for class 'gamfit'
print(x, ...)

## S3 method for class 'gamfit'
summary(object, ...)

## S3 method for class 'gamfit'
fitted(object, ...)

## S3 method for class 'gamfit'
plot(x, pick = 1, pcol, lcol, pcex, lwd, labcex,
  title = TRUE, nlevels = 10, levels, xlab, ylab, ...)

gamfit_sens(x, env, nrep = 99, perc = 5, ...)

## S3 method for class 'gamfit_sens'
plot(r, stat = "r2", pltype = "joy", ...)

## S3 method for class 'gamfit_sens'
summary(r, x, ...)

Arguments

ord	ordination model from vegan
env	environmental matrix with variables to evaluate
...	additional arguments passed to gam
x, object	object of class gamfit
pick	numeric, the column number of environmental variable to plot as a nonlinear regression surface in the ordination space
pcol, lcol	point and line colors
pcex, lwd	point size and line width
title	logical, add variable name as title? (default=TRUE)
nrep	number of repetitions for sensitivity analysis (default=99)
perc	numeric, percent noise to add to env variables for sensitivity analysis
r	object of class gamfit_sens, from sensitivity analysis
stat	one of c('r2','pval') to plot from sensitivity analysis
pltype	one of c('heat','joy') for a heatmap or joyplot, respectively

Details

Nonlinear regression of each environmental variable in turn on the site scores resulting from NMDS ordination. Returns p-values for the joint smooth term, and adjusted-R2 expressing the strength of (nonlinear) relationship between each environmental variable and the ordination. Print, summary and plot methods exist. Sensitivity analysis currently exists but needs refinement.

Value

Object of class gamfit or gamfit_sens for those functions respectively, or plots to device.

See Also

gam for the internal fitting function, ordisurf for a similar procedure that only admits variables one-at-a-time, and envfit for a linear alternative. Plotting methods for sensitivity analysis follow plot_joy and plot_heatmap.

Examples

data(veg)
spe <- veg$spe
env <- veg$env
m <- vegan::metaMDS(vegan::vegdist(spe), trace=0)
g <- gamfit(m, env)
g
plot(g, pick=10)

# sensitivity analysis
r <- gamfit_sens(g, env)
plot(r, pltype='heat')
plot(r, pltype='joy')
(MAE <- summary(r, g))
plot(MAE, xaxt='n', las=1)
axis(1, 1:NCOL(env), names(env), cex.axis=0.7)
sapply(data.frame(r[,,1]<0.05),sum) / 99 # proportion 'significant'

phytomosaic/foggy documentation built on Nov. 5, 2019, 12:20 a.m.