VCA (version 1.3.1)

stepwiseVCA: Bottom-Up Step-Wise VCA-Analysis of the Complete Dataset.

Description

Function performs step-wise VCA-analysis on a fitted VCA-object by leaving out N-1 to 0 top-level variance components (VC).

Usage

stepwiseVCA(obj, VarVC = FALSE, VarVC.method = c("scm", "gb"))

Arguments

obj

(VCA) object representing the complete analysis

VarVC

(logical) TRUE = estimate complete variance-covariance matrix of variance components and include corresponding sub-matrices in result-objects of the step-wise analysis. This allows to cosntruct confidence intervals on all VCs via function VCAinference

VarVC.method

(character) string specifying the algorithm to be used for estimating variance-covariance matrix of VCs (see anovaMM for details).

Value

(list) of (simplified) 'VCA' objects representing analysis-result of sub-models

Details

This function uses the complete data to quantify sub-sets of variance components. In each step the current total variance is estimated by subtracting the sum of all left-out VCs from the total variance of the initial VCA object. Doing this guarantees that the contribution to the total variance which is due to left-out VCs is accounted for, i.e. it is estimated but not included/reported. The degrees of freedom (DFs) the emerging total variances of sub-sets are estimated using the Satterthwaite approximation. This is achieved by extracting the corresponding sub-matrix from the coefficient matrix $C$ of the 'VCA' object, the sub-vector of ANOVA mean squares, and the sub-vector of degrees of freedom and calling function SattDF method="total". This step-wise procedure starts one-level above error (repeatability) and ends at the level of the upper-most VC.

Examples

## Not run: 
# data(VCAdata1)
# datS7L1 <- VCAdata1[VCAdata1$sample == 7 & VCAdata1$lot == 1, ]
# fit0 <- anovaVCA(y~device/day/run, datS7L1, MME=TRUE)
# 
# # complete VCA-analysis result
# fit0
# 
# # perform step-wise (bottom-up) VCA-analyses
# sw.res <- stepwiseVCA(fit0, VarVC=TRUE)
# sw.res
# 
# # get CIs on intermediate precision
# VCAinference(sw.res[["device:day"]], VarVC=TRUE)
# ## End(Not run)