error in biv model with definition variable

5 replies [Last post]
Nienke's picture
Offline
Joined: 01/22/2013

Hi,
I am trying to fit a biv model (2 continuous phenotypes) with sex as definition variable to test for mean differences, however I get the same error over and over again:

Error: A definition variable has been declared in model 'Chol' that does not contain a data set

I cannot find where it goes wrong, I suspect it's somewhere in the matrices declared to store linear coefficients for covariate:

grandMean <- mxMatrix(type="Full", nrow=1, ncol=nphen, free = TRUE, values=c(2700, 19), label=c("mean1","mean2"), name="Mean")
B_Sex <- mxMatrix(type="Full", nrow=ndef, ncol=nvar, free=TRUE, values=c(900,2.7), label=(rep(c("bphen1","bphen2"), 2)), name="bSex" )
defSex <- mxMatrix(type="Full", nrow=ndef, ncol=nvar, free=FALSE, labels=(rep(c("data.Sex1","data.Sex2"),each=2)), name="Sex")
SexR <- mxAlgebra(bSex * Sex, name="SexR")
expMean <- mxAlgebra(name="expMean", expression= cbind(Mean, Mean) + SexR)

Any help would be very much appreciated!

Regards,
Nienke

tbates's picture
Offline
Joined: 07/31/2009
definition variable declared in model that does not contain data

The error appears to be saying that it wants you to add a dataset to the model, so that it can find the variable?
Try adding something like this to your model:

	mxData(myData, type="raw")

PS, the tiny fragment of script you posted doesn't help, as it can't be generating the error

randMean <- mxMatrix(type="Full", nrow=1, ncol=nphen, free = TRUE, values=c(2700, 19), label=c("mean1","mean2"), name="Mean")
B_Sex <- mxMatrix(type="Full", nrow=ndef, ncol=nvar, free=TRUE, values=c(900,2.7), label=(rep(c("bphen1","bphen2"), 2)), name="bSex" )
defSex <- mxMatrix(type="Full", nrow=ndef, ncol=nvar, free=FALSE, labels=(rep(c("data.Sex1","data.Sex2"),each=2)), name="Sex")
SexR <- mxAlgebra(bSex * Sex, name="SexR")
expMean <- mxAlgebra(name="expMean", expression= cbind(Mean, Mean) + SexR)

James Sherlock's picture
Offline
Joined: 06/03/2014
definition in model that does not contain data

I'm having a similar error with age as a covariate in a trivariate model but definitely have specified mxData

Error: A definition variable has been declared in model 'CholACE' that does not contain a data set

The full script is below
data <- read.csv('Standardized Disgust Data.csv', na.strings='-99')

head(data)
describe(data)

data$twin1age <- data$age_NEW.1
data$twin2age <- data$age_NEW.2
data$sibage <- data$sib1age
data$moral1 <- data$ZDisgustMoral.1
data$moral2 <- data$ZDisgustMoral.2
data$sibdm <- data$Zsib1dm
data$path1 <- data$ZDisgustPathogen.1
data$path2 <- data$ZDisgustPathogen.2
data$sibdp <- data$Zsib1dp
data$sexual1 <- data$ZDisgustSexual.1
data$sexual2 <- data$ZDisgustSexual.2
data$sibds <- data$Zsib1ds
data$zyg <- data$Corrected_zygosity
head(data)

# Select Variables for Analysis
Vars <- c('moral','path','sexual')
nv <- 3 # number of variables
nsib <- 3
ntv <- nv*nsib # number of total variables

selVars <- c('moral1','path1','sexual1',
'moral2','path2','sexual2',
'sibdm','sibdp','sibds')

defVars <-c('twin1age','twin2age','sibage')
useVars <- c(selVars,defVars)

#Subset data for testing
mzdata <- subset(data, zyg==3, useVars)
dzdata <- subset(data, zyg==4, useVars)
describe(mzdata, skew=F)
describe(dzdata, skew=F)
dim(mzdata)
dim(dzdata)
cov(mzdata,use="complete")
cov(dzdata,use="complete")
cor(mzdata,use="complete")
cor(dzdata,use="complete")
# ------------------------------------------------------------------------------
# ------------------------------------------------------------------------------
# Cholesky Decomposition ACE Model
# ------------------------------------------------------------------------------
# Set Starting Values
svMe <- c(.01,.01, .01) # start value for means
svPa <- valDiag(nv,.6) # start values for parameters on diagonal
lbPa <- valLUDiag(nv,.0001,-10,NA) # lower bounds for parameters on diagonal

# Matrices declared to store a, c, and e Path Coefficients
pathA <- mxMatrix( type="Lower", nrow=nv, ncol=nv, free=TRUE,
values=svPa, labels=labLower("a",nv), lbound=lbPa, name="a" )
pathC <- mxMatrix( type="Lower", nrow=nv, ncol=nv, free=TRUE,
values=svPa, labels=labLower("c",nv), lbound=lbPa, name="c" )
pathE <- mxMatrix( type="Lower", nrow=nv, ncol=nv, free=TRUE,
values=svPa, labels=labLower("e",nv), lbound=lbPa, name="e" )

# Matrices generated to hold A, C, and E computed Variance Components
covA <- mxAlgebra( expression=a %*% t(a), name="A" )
covC <- mxAlgebra( expression=c %*% t(c), name="C" )
covE <- mxAlgebra( expression=e %*% t(e), name="E" )

# Algebra to compute total variances and standard deviations (diagonal only)
covP <- mxAlgebra( expression=A+C+E, name="V" )
matI <- mxMatrix( type="Iden", nrow=nv, ncol=nv, name="I")
invSD <- mxAlgebra( expression=solve(sqrt(I*V)), name="iSD")

# Algebra for expected Mean and Variance/Covariance Matrices in MZ & DZ twins
# Algebra for expected Mean Matrices in MZ & DZ twins
defAge <- mxMatrix(type="Full", nrow=1,ncol=ntv,free=FALSE,
labels=c('data.twin1age','data.twin2age','data.sibage'),name='Age')

pathB <- mxMatrix( type="Full", nrow=1, ncol=1, free=TRUE, values= .01,
label="b11", name="b")

meanG <- mxMatrix( type="Full", nrow=1, ncol=ntv, free=TRUE,
values=svMe, labels=labFull("me",1,nv), name="meanG" )

expMean <- mxAlgebra( meanG + (b%x%Age), name="expMean" )

covMZ <- mxAlgebra( expression= rbind( cbind(V , A+C, .5%x%A+C),
cbind(A+C, V, .5%x%A+C),
cbind(.5%x%A+C, .5%x%A+C, V)),
name="expCovMZ" )

covDZ <- mxAlgebra( expression= rbind( cbind(V , .5%x%A+C, .5%x%A+C),
cbind(.5%x%A+C, V, .5%x%A+C),
cbind(.5%x%A+C, .5%x%A+C, V)),
name="expCovDZ" )

# Data objects for Multiple Groups
dataMZ <- mxData( observed=mzdata, type="raw" )
dataDZ <- mxData( observed=dzdata, type="raw" )

# Objective objects for Multiple Groups
objMZ <- mxFIMLObjective( covariance="expCovMZ", means="expMean", dimnames=useVars )
objDZ <- mxFIMLObjective( covariance="expCovDZ", means="expMean", dimnames=useVars )

# Combine Groups
pars <- list( pathA, pathC, pathE, covA, covC, covE, covP, matI, invSD, defAge, pathB, meanG, expMean )
modelMZ <- mxModel( pars, covMZ, dataMZ, objMZ, name="MZ" )
modelDZ <- mxModel( pars, covDZ, dataDZ, objDZ, name="DZ" )
minus2ll <- mxAlgebra( expression=MZ.objective + DZ.objective, name="m2LL" )
obj <- mxAlgebraObjective( "m2LL" )
CholAceModel <- mxModel( "CholACE", pars, modelMZ, modelDZ, minus2ll, obj )

# ------------------------------------------------------------------------------
# RUN GENETIC MODEL

# Run Cholesky Decomposition ACE model
CholAceFit <- mxRun(CholAceModel)

RobK's picture
Offline
Joined: 04/19/2011
Should be simple to fix

You are not the first user to run into this problem. Try moving the matrix for the definition variables, defAge, out of pars and into the submodel being declared.
That should fix the problem. Or, you could just remove pars from the supermodel. So, change
CholAceModel <- mxModel( "CholACE", pars, modelMZ, modelDZ, minus2ll, obj )
to
CholAceModel <- mxModel( "CholACE", modelMZ, modelDZ, minus2ll, obj )
as well. The cause is the presence of defAge in the supermodel, CholAceModel, which does not contain any data of its own.

Sorry if my prior edits of this post were confusing... :-/

James Sherlock's picture
Offline
Joined: 06/03/2014
Fantastic, thank you!

Fantastic, thank you!

Nienke's picture
Offline
Joined: 01/22/2013
Sorry, you're right, think

Sorry, you're right, think I've found the solution, however next time I'll post more of the script :)