forsearch (version 6.4.0)

forsearch_lme: Create Statistics Of Forward Search For a Linear Mixed Effects Database

Description

Prepares summary statistics at each stage of forward search for subsequent plotting. Forward search is conducted in four steps: Step 0 to set up accounting for group structure, Step 1 to identify minimal set of observations to estimate unknown fixed parameters, Step 2 to identify the order of the remaining observations, and a final stage to extract the intermediate statistics based on increasing sample size.

Usage

forsearch_lme(fixedform, nofactform, alldata, randomform, groupname, randfactnames=NULL,
   initial.sample=1000, skip.step1=NULL, unblinded=TRUE, begin.diagnose = 100, 
   incCont=FALSE, verbose = TRUE)

Value

LIST

Number of observations in Step 1

Number of observations included in Step 1

Step 1 observation numbers

Observation numbers useful in skipping step 1

Rows by outer subgroup

List of row numbers, by outer subgroup

Rows by outer-inner subgroups

List of row numbers, by outer-inner subgroup

Rows in stage

Observation numbers of rows included at each stage

Sigma

Estimate of random error at final stage; used to standardize all residuals

Standardized residuals

Matrix of errors at each stage

Fixed parameter estimates

Matrix of parameter estimates at each stage

Random parameter estimates

Matrix of parameter estimates at each stage

Leverage

Matrix of leverage of each observation at each stage

Modified Cook distance

Estimate of sum of squared changes in parameter estimates at each stage

Dims

Dims from fit of lme function

t statistics

t statistics for each fixed parameter

Fit statistics

AIC, BIC, and log likelihood

Call

Call to this function

Arguments

fixedform

2-sided formula for fixed effects

nofactform

2-sided formula for fixed effects, omitting factors

alldata

data frame, first column of which must be "Observation"

randomform

1-sided formula for random effects

groupname

Quoted name of group variable in randomform. Nested grouping not permitted in this version

randfactnames

Vector of quoted names of random factor variables

initial.sample

Number of observations in Step 1 of forward search

skip.step1

NULL or a vector of integers for observations to be included in Step 1

unblinded

TRUE causes printing of presumed analysis structure

begin.diagnose

Numeric indicator of place in coding to begin printing diagnostic information. 0 prints all information, 100 prints none.

incCont

Logical. Currently ignored

verbose

TRUE causes function identifier to display before and after run

Author

William R. Fairweather

Details

data will be grouped within the function, regardless of initial layout. Step 2 is determined by the results of Step 1, which itself is random. So, it is possible to reproduce the entire run by using the skip.step1 argument. Variables in the randomform formula must be character variables, but *not* factors

References

Atkinson, A and M Riani. Robust Diagnostic Regression Analysis, Springer, New York, 2000. Pinheiro, JC and DM Bates. Mixed-Effects Models in S and S-Plus, Springer, New York, 2000. https://CRAN.R-project.org/package=nlme

Examples

if (FALSE) {
# Multiple regression in grouped data
Observation <- 1:160
y <- runif(160)
x1 <- runif(160)
x2 <- runif(160)
x3 <- runif(160)
group <- rep(c("G1","G2"),each=80)
lmetest1 <- data.frame(Observation,y,x1,x2,x3,group)
forsearch_lme(fixedform=y~x1+x2+x3, nofactform=y~x1+x2+x3, alldata=lmetest1, 
   randomform= ~1|group, groupname="group", initial.sample=200)

# Analysis of variance in grouped data
Observation <- 1:60
y <- runif(60)
AN1 <- as.factor(c(rep("A1",5),rep("A2",5),rep("A3",5)))
AN1 <- c(AN1,AN1,AN1,AN1)
AN2 <- as.factor(c(rep("B1",15),rep("B2",15)))
AN2 <- c(AN2,AN2)
group <- rep(c("G1","G2"),each=30)
lmetest2 <- data.frame(Observation,y,AN1,AN2,group)
forsearch_lme(fixedform=y~AN1*AN2, nofactform=y~1, alldata=lmetest2, 
   randomform= ~1|group, groupname="group",initial.sample=500)

# Analysis of covariance in grouped data

Observation <- 1:120
y <- runif(120)
AN1 <- as.factor(c(rep("A1",10),rep("A2",10),rep("A3",10),rep("A4",10)))
AN1 <- c(AN1,AN1,AN1)
AN2 <- as.factor(c(rep("B1",5),rep("B2",5)))
AN2 <- c(AN2,AN2,AN2,AN2,AN2,AN2)
AN2 <- c(AN2,AN2)
COV <- runif(120)
group <- rep(c("G1","G2"),each=30)
group <- c(group,group)
lmetest3 <- data.frame(Observation,y,AN1,AN2,COV,group)
test3<-forsearch_lme(fixedform=y~AN1*AN2+COV,nofactform=y~COV,alldata=lmetest3,
        randomform= ~ 1 | group,groupname="group",initial.sample=500,
        begin.diagnose=100)
}