mratios (version 1.4.2)

sci.ratio.gen: Simultaneous confidence intervals for ratios of coefficients in the general linear model

Description

Constructs simultaneous confidence intervals for multiple ratios of linear combinations of coefficients in the general linear model.

Usage

sci.ratio.gen(Y, X, Num.Contrast, Den.Contrast,
 alternative = "two.sided", conf.level = 0.95,
 method="Plug")

Arguments

Y

A numerical vector, containing the values of the response variable

X

A design matrix for the the linear model, defining the parameters to be estimated, must have same number of rows as Y

Num.Contrast

Numerator contrast matrix

Den.Contrast

Denominator contrast matrix

alternative

one of "two.sided", "less", or "greater"

conf.level

simultaneous confidence levels

method

character string, specifying the method for confidence interval calculation:

• "Plug": Plug-in of ratio estimates in the correlation matrix of the multivariate t distribution. This method is the default.

• "Bonf": Simple Bonferroni-adjustment of Fieller confidence intervals for the ratios

• "MtI": Sidak or Slepian- adjustment for two-sided and one-sided confidence intervals, respectively

• "Unadj": Unadjusted Fieller confidence intervals for the ratios (i.e. with comparisonwise confidence level = conf.level)

Value

A list containing

estimate

point estimates for the ratios

CorrMat.est

estimates of the correlation matrix (for the plug-in approach)

Num.Contrast

matrix of contrasts used for the numerator of ratios

Den.Contrast

matrix of contrasts used for the denominator of ratios

conf.int

confidence interval estimates of the ratios

Y

response vector

X

design matrix

fit

the model fit, an object of class "lm"

and some further input arguments, to be passed to print and summary functions.

Details

Given a general linear model, the interest is in simultaneous confidence intervals for several ratios of linear combinations of the coefficients in the model. It is assumed that the responses are normally distributed with homogeneous variances. In this problem, the joint distribution of the likelihood ratio statistics has a multivariate t-distribution the correlation matrix of which depends on the unknown ratios. This means that the critical point needed for CI calculations also depends on the ratios. There are various methods of dealing with this problem (for example, see Dilba et al., 2006). The methods include (i) the unadjusted intervals (Fieller confidence intervals without multiplicity adjustments), (ii) Bonferroni (Fieller intervals with simple Bonferroni adjustments), (iii) MtI (a method based on Sidak and Slepian inequalities for two- and one-sided confidence intervals, respectively), and (iv) plug-in (plugging the maximum likelihood estimates of the ratios in the unknown correlation matrix). The MtI method consists of replacing the unknown correlation matrix by an identity matrix of the same dimension.

Applications include relative potency estimations in multiple parallel line or slope-ratio assays. Users need to define the design matrix of the linear model and the corresponding contrast matrices in an appropriate way.

References

Dilba, G., Bretz, F., and Guiard, V. (2006). Simultaneous confidence sets and confidence intervals for multiple ratios. Journal of Statistical Planning and Inference 136, 2640-2658.

See Also

glht(multcomp) for multiple comparisons of parameters from lm, glm,..., sci.ratio for confidence intervals for ratios of means in a one-way-layout, simtest.ratio for simultaneous tests for ratios of means in a one-way-layout, plot.sci.ratio for plotting the confidence intervals.

Examples

# NOT RUN {
################################################

# Slope-ratio assay on data from Jensen(1989),
# Biometrical Journal 31, 841-853.

# Definition of the vector of responses and
# the design matrix can be done directly as
# follows:

Y0 <- c(1.3, 1.7, 2.4, 2.7, 3.6, 3.6, 4.7, 5.0, 6.1, 6.3)
Y1 <- c(2.8, 2.9, 4.1, 3.7, 5.5, 5.5, 6.4, 6.7)
Y2 <- c(2.2, 2.1, 3.2, 3.2, 3.8, 3.9, 4.7, 4.9)
Y3 <- c(2.3, 2.3, 3.2, 3.0, 4.2, 4.2, 4.6, 5.1)
Y <- c(Y0,Y1,Y2,Y3) # the response vector

xi <- rep(1,34)
x0 <- c(0,0, gl(4,2),rep(0,8*3))
x1 <- c(rep(0,10),gl(4,2), rep(0,8*2))
x2 <- c(rep(0,18),gl(4,2), rep(0,8))
x3 <- c(rep(0,26),gl(4,2))

X <- cbind(xi,x0,x1,x2,x3) # the design matrix

# Have a look at the response vector:
Y

# and the design matrix:

X

# Internally in sci.ratio.gen, the following model is fitted
 
Fiti <- lm(Y ~ X - 1)
Fiti
summary(Fiti)

# In this problem, interest is simultaneous estimation of
# the ratios of slopes relative to the slope of the standard
# treatment. Therefore, the appropriate contrast matrices are:

Num.Contrast <- matrix(c(0,0,1,0,0,
                         0,0,0,1,0,
                         0,0,0,0,1),nrow=3,byrow=TRUE)
Den.Contrast <- matrix(c(0,1,0,0,0,
                         0,1,0,0,0,
                         0,1,0,0,0),nrow=3,byrow=TRUE)



SlopeRatioCI <- sci.ratio.gen(Y=Y, X=X,
 Num.Contrast=Num.Contrast, Den.Contrast=Den.Contrast)

SlopeRatioCI

# Further details of the fitted model and the contrasts used:

summary(SlopeRatioCI)

plot(SlopeRatioCI)


#########################################################

# }
# NOT RUN {
# If one starts with a dataframe, the function model.matrix
# can be used to create the design matrix:

data(SRAssay)
SRAssay

# Create the design matrix using model.matrix

X <- model.matrix(Response~Treatment:Dose, data=SRAssay)
Response <- SRAssay[,"Response"]

# The response vector and the design matrix are now:

X
Response

# The following coefficients result from fitting this model:

lm(Response~0+X)

# The same contrasts as above are used:

Num.Contrast <- matrix(c(0,0,1,0,0,
                         0,0,0,1,0,
                         0,0,0,0,1),nrow=3,byrow=TRUE)
Den.Contrast <- matrix(c(0,1,0,0,0,
                         0,1,0,0,0,
                         0,1,0,0,0),nrow=3,byrow=TRUE)

summary(sci.ratio.gen(Y=Response, X=X, Num.Contrast, Den.Contrast))

# }