get_cor: Compute the prediction accuracy from Cross Validition

In cheuerde/cpgen: Parallel Genomic Evaluations

Compute the prediction accuracy from Cross Validition

Description

Takes a matrix of predictions returned by get_pred, a list of masked phenotypes returned by cCV and the original phenotype vector and returns the correlation between predicted and observed values

Usage

get_cor(predictions,cv_pheno,y)

Arguments

predictions	Prediction matrix returned by get_pred
cv_pheno	List of masked phenotypes returned by cCV
y	Original unmasked phenotype vector that has been used in cCV

Value

Numeric scalar - Mean prediction accuracy measured as correlation between predicted and observed phenotypes

See Also

clmm, get_pred, cCV

Examples

### Running a 4-fold cross-validation with one repetition:

# generate random data
rand_data(100,500)

### compute the list of masked phenotype-vectors for CV
y_CV <- cCV(y,fold=4,reps=1)


### Cross Validation using GBLUP
G.A <- cgrm.A(M,lambda=0.01)


### generate the list of design matrices for clmm
Z_list = list(t(chol(G.A)))

### specify options
h2 = 0.3
scale = unlist(lapply(y_CV,function(x)var(x,na.rm=TRUE))) * h2
df = rep(5,length(y_CV))
par_random = list(list(method="ridge",scale=scale,df=df))

### run 
fit <- clmm(y_CV, Z=Z_list, par_random=par_random, niter=5000, burnin=2500)

### inspect results
str(fit)

### obtain predictions
pred <- get_pred(fit)

### prediction accuracy
get_cor(pred,y_CV,y)

cheuerde/cpgen documentation built on July 27, 2023, 11:34 a.m.