nlsregression: nlsregression

In pik-piam/regressionworlddata: Regression analysis based on global datasets

nlsregression

Description

Creates regression parameter estimates and plots with any function you want that has no more than two independent variables

Usage

nlsregression(
  func,
  y,
  x,
  z = NULL,
  startvalues = NULL,
  weight = NULL,
  weighting = TRUE,
  xlab = NULL,
  ylab = "y",
  header = NULL,
  z_plot_lines = NULL,
  weightcolorpoints = TRUE,
  x_log10 = FALSE,
  toPlot = "all",
  plot_x_function = "ignore",
  regressioncolor = "blue",
  weight_threshold = NULL,
  crossvalid = NULL,
  ...
)

Arguments

func	function that shall be fitted. Function should contain the dependent variable y and and the independent variable x, eventually a second independent variable z. All other unknowns are treated as parameters that are estimated.
y	dependent variable,vector
x	independent variable,vector
z	optional independent variable,vector
startvalues	the optimization algorithm may require starting values for the fitting procedure. provide them in a list with the parameter names: e.g. list(a=3,b=2)
weight	optional weight,vector
weighting	if weighting is TRUE, the fit will minimize the weighted residuals
xlab	name of x axis in plot
ylab	name of y axis in plot
header	plot function main argument
z_plot_lines	vector>1 of values for z you want to be plotted into the graph
weightcolorpoints	if TRUE, the points are clustered into three quantiles according to their weight and coloured lighter for low weights.
x_log10	allows log10 scale for X axis if set to TRUE. Only changes the picture, not the regression!
toPlot	"all", "frame" (axis etc), "observations" (points), "regressionline" (line), "infos" (parameters, R2)
plot_x_function	depreciated, please do not enter into function call.
regressioncolor	color of regression line and paramter text
weight_threshold	if numeric, all countries below this threshold will be excluded (e.g. to exclude minor islands)
crossvalid	vector with boolean values, indicating which data should be excluded from sampling and rather be used for validation
...	will be passed on to function nls

Value

A nice picture and regression parameters or eventually some errors.

Author(s)

Benjamin Leon Bodirsky, Susanne Rolinski, Xiaoxi Wang

Examples

## Not run: 
x=1:10
y=(1:10)^2+1
z=c(10:1)

# one independent variable
nlsregression(func=y~a*x+b,y=y,x=x,startvalues=list(a=1,b=1))
# two independent variables
nlsregression(func=y~a*x^1.1+b*z+c*x,y=y,x=x,z=z,startvalues=list(a=1,b=1,c=0))
# no fit because residuals are zero (excluded from the nls makers due
 to statistical reasons)
nlsregression(func=y~x^a+b,y=y,x=x,z=z,startvalues=list(a=1,b=1,c=0))

DNase1 <- subset(DNase, Run == 1)
DNase1$sets<- c(rep(1,8),rep(2,8))
nlsregression(func=y~a*x+b,y=DNase1$density,x=DNase1$conc,startvalues=list(a=1,b=1))
nlsregression(func=y~a*x+b*z,y=DNase1$density,x=DNase1$conc,z=DNase1$sets,
startvalues=list(a=0.1344,b=0.2597))
nlsregression(func=y~a*x+b*z,y=DNase1$density,x=DNase1$conc,z=DNase1$sets,
startvalues=list(a=0.1344,b=0.2597),plot_x_function=log)

## End(Not run)

pik-piam/regressionworlddata documentation built on June 6, 2024, 11:08 p.m.