sis: Sure Independent Screening
In YinanZheng/PGS: Penalized GEE with Grid Search

Description Usage Arguments Value See Also Examples

View source: R/sis.R

sis is used to conduct sure independent screening across high-dimensional variables.

1
2
3

sis(y.vect, id.vect = NULL, M, COV = NULL, method = c("LMM", "GEE",
  "MLR"), corstr = "ar1", parallel = TRUE, ncore = detectCores(),
  write = FALSE)

`y.vect`	a vector of dependent variable.
`id.vect`	a vector of subjuect ID.
`M`	a data frame or matrix of genomic dataset. Rows represent samples, columns represent variables.
`COV`	a data frame or matrix of covariates dataset.
`method`	a character string specifying fitting method. For data contains >= 2 repeated measures, linear mixed-effect model (`"LMM"`) and generalized estimation equation (`"GEE"`) method are supported. For single measurement data, simple/multiple linear regression (`"MLR"`) is available. Default = `"LMM"`.
`corstr`	a character string specifying the correlation structure when `method = "GEE"`. The following are permitted: '"independence"', '"exchangeable"', '"ar1"', '"unstructured"' and '"userdefined"'. Default = "ar1". (see `geeglm`)
`parallel`	logical. Enable parallel computing feature? Default = `TRUE`.
`ncore`	number of cores to run parallel. Effective when paralle = `TRUE`. By default, max number of cores will be used.
`write`	logical. Export screening results to csv file in the working directory if `TRUE`. Defaul = `FALSE`.

sure independent screening results in a sis.obj object.

see pgsfit using the results from sis as input to run PGS; see sis.obj for class methods; see lmer, geeglm, and lm for more details on "LMM","GEE", and "MLR" methods, respectively.

### Dataset preview
BJdata()

### Convert binary variables into factor type 
BJlung$gender = factor(BJlung$gender)
BJlung$heat = factor(BJlung$heat)
BJlung$cigwear = factor(BJlung$cigwear)

### Merge miRNA and lung function dataset
BJdata <- merge(BJmirna, BJlung, by=c("SID","WD"))

### Data must be sorted by study subject ID and multiple measurements indicator
BJdata <- BJdata[with(BJdata, order(SID, WD)), ]

### Extract dependent variable (lung function)
y.vect<-BJdata$FEV1

### Extract subjuect ID variable indicating repeated measures             
id.vect<-BJdata$SID        

### Extract microRNA data matrix   
M<-BJdata[,3:168]   

### Extract covariate data matrix          
COV<-BJdata[,170:179]
           
### LMM sure independent screening results
sis_LMM_par = sis(y.vect, id.vect, M, COV, method = "LMM")

### GEE sure independent screening results
sis_GEE_par = sis(y.vect, id.vect, M, COV, method = "GEE")

### Save the full site-by-site testing results into a csv file in current working directory
sis_LMM_par = sis(y.vect, id.vect, M, COV, method = "LMM", write = T)

sis_LMM_par        # print summary of sure independent screening results
plot(sis_LMM_par)  # plot histogram of raw p-values and Q-Q plot
coef(sis_LMM_par)  # return coefficients from sure independent screening results

#For more information, please visit: https://github.com/YinanZheng/PGS/wiki/Example:-miRNA-expression-and-lung-function