leapp: latent effect adjustment after primary projection

In leapp: Latent Effect Adjustment After Primary Projection

latent effect adjustment after primary projection

Description

Adjust for latent factors and conduct multiple hypotheses testing from gene expression data using the algorithm of Sun,Zhang and Owen (2011). Number of latent factors can be chosen by Buja and Eyuboglu (1992).

Usage

leapp (data,pred.prim,pred.covar,
         O = NULL, num.fac = "buja", method = "hard", sparse = TRUE,
         centered = FALSE, verbose = FALSE, perm.num = 50, 
         TOL = 1e-4, length.out = 50)

Arguments

data	An N genes by n arrays matrix of expression data
pred.prim	An n by 1 primary predictor
pred.covar	An n by s known covariate matrix not of primary interest
O	An n by n rotation matrix such that O pred.prim = (1, 0,...,0)
num.fac	A numeric or string, number of latent factors chosen. it has default value "buja" which uses Buja and Eyuboglu (1992) to pick the number of factors
method	A string which takes values in ("hard","soft"). "hard": hard thresholding in the IPOD algorithm; "soft": soft thresholding in the IPOD algorithm
sparse	A logical value, if TRUE, the signal is sparse and the proportion of non-null genes is small, use IPOD algorithm in Owen and She (2010) to enforce sparsity. If FALSE, the signal is not sparse, use ridge type penalty to carry out the inference as in Sun,Zhang, Owen (2011). Default to TRUE
centered	A logical value, indicates whether the data has been centered at zero, default to FALSE
verbose	A logical value, if TRUE, will print much information as algorithm proceeds, default to FALSE
perm.num	A numeric, number of permutation performed using algorithm of Buja and Eyuboglu (1992), default to 50
TOL	A numeric, convergence tolerance level, default to 1e-4
length.out	A numeric, number of candidate tuning parameter lambda under consideration for further modified BIC model selection, default to 50.

Details

The data for test i should be in the ith row of data. If the rotation matrix O is set to NULL, the function will compute one rotation from primary predictor pred.prim.

Value

p	A vector of p-values one for each row of data.
vest	An n by num.fac matrix, estimated latent factors
uest	An N by num.fac matrix, estimated latent loadings
gamma	An N by 1 vector, estimated primary effect
sigma	An N by 1 vector, estimated noise standard deviation one for each row of data

Author(s)

Yunting Sun [e-mail], Nancy R.Zhang [e-mail], Art B.Owen [e-mail]

Examples

  ## Not run: 
  ## Load data
  data(simdat)
  
  
    
  #Calculate the p-values
  p <- leapp(simdat$data,pred.prim = simdat$g,method = "hard")$p
  auc <- FindAUC(p, which(simdat$gamma!=0))
  
  
 
## End(Not run)

leapp documentation built on June 20, 2022, 1:05 a.m.