curvetest (version 2.2)

curvetest: Test Equality of Curves with Homoscedastic or Heteroscedastic Errors.

Description

Main test routine for formally testing the equality of two curves represented by discrete data points over a domain with homoscedacity or heteroscedacity errors. curvetest is the wrapper of curvetest.raw.

Usage

curvetest(formula, data1 = NULL, data2 = NULL, equal.var = TRUE, alpha = 0.5, bw = NULL, plotit = TRUE, conf.level = 0.05, kernel = c("Triangle", "Gaussian", "Trio", "Uniform", "Triweight", "Epanechnikov", "Quartic"), nn = 100, myx = NULL, bcorrect = "simple",...)
curvetest.raw(fits1, fits2, equal.var, conf.level, plotit)
"print"(x,...)

Arguments

formula

A formula to the data set such as y~x.

data1

A data frame of 2 columns representing the underlying curve 1. The column names must agree with the names in formula.

data2

A data frame for curve 2. If it is NULL, the test is whether curve 1 is statistically equal to 0 over the defining domain.

equal.var

Whether the variances are equal. Default to TRUE.

alpha

Smoothing parameter.

bw

Window bandwidth for both curves.

plotit

WHether plot the fitted curves or not. Default: FALSE.

conf.level

The confidence level to claim the curves are different. Default: 0.05.

kernel

One of the kernel functions to use to fit the curves. Must be one of "Triangle", "Gaussian", "Trio","Uniform", "Triweight", "Epanechnikov", "Quartic". partial match is allowed.

nn

Number of data points in the test domain to calculate the curve values.

myx

x-values in the test domain to calculate the curve values. If it is specified, nn will be suppressed.

bcorrect

Boundary correction method. Right now, except for 'none', meaning no corrections, the only other opton is 'simple'.

...

When plotit is true, plot parameters can be specified such as pch, lty, col etc.

fits1,fits2

The fitted results from fitting the first or second curve by curvefit procedure.

x

Test results from curvetest.

Value

Statistic,p

Test statistic and p value of testing whether $f_1(x)==f_2(x)$ or $f_1(x)==0$.

eDF

Estimated degree of freedom of the fitting.

equal.var

The model specification of whether the two variances are equal or not

esigma1,esigma2

Estimated varaince of the fitted curves.

k0

The calculated value from the tube. See detail in paper.

fits1,fits2

Objects of class 'curvefit' from curvefit ruotine.

Details

The algorithm works by first fitting the curves using local regression method specified by formula on data1 and/or data 2 with smoothing parameters specified in the function calls. Then it will test on the fitted curve 1 and curve 2 to see if they are statistically equal or not.

References

Zhongfa Zhang, et al: Test Equality of Curves with Homoscedastic or Heteroscedastic Errors. To appear

See Also

curvefit, print.curvetest, plot.curvetest

Examples

    n1=150; n2=155 ##numbers of data points for the two curves. 
    f1<-f2<-function(x){x*(1-x)+sin(2*pi*x)}; ##True functions.
    x1=seq(0,1, length=n1);
    x2=seq(0, 1, length=n2);
    y1=f1(x1)+rnorm(n1, 0, 0.2)
    y2=f2(x2)+rnorm(n2, 0, 0.2) ###Measured data for the two curves with noises. 
    curvetest(y~x,data.frame(x=x1,y=y1), data.frame(x=x2,y=y2), alpha =  0.7, equal.var=TRUE,plotit=TRUE)