R/describe.by.R

In psych: Procedures for Psychological, Psychometric, and Personality Research

#modified March 4, 2009 for matrix output
#and yet again August 1 to make it actually work!
#modified May 26, 2014 to add the ability to specify group by name or location
"describe.by" <-
function (x,group=NULL,mat=FALSE,type=3,...) {               #data are x, grouping variable is group
.Deprecated("describeBy", msg = "describe.by is deprecated.  Please use the describeBy function")
  if(inherits(x,"formula")) {ps <- fparse(x) 
    x <- ps$y
group <- ps$x} 
answer <- describeBy(x=x,group=group,mat=mat,type=type,...)
return(answer)}

#July 9, 2020  added the ability to do formula input
"describeBy" <-
function (x,group=NULL,mat=FALSE,type=3,digits=15,data=NULL,...) {               #data are x, grouping variable is group
   cl <- match.call()
   if(inherits(x,"formula")) {ps <- fparse(x) 
   if(missing(data)) {x <- get(ps$y)
   group <- x[ps$x]} else {x <- data[ps$y]
   group <- data[ps$x]}
} 
 if(is.null(group)) {answer <- describe(x,type=type)
   warning("no grouping variable requested")} else {
   x <- char2numeric(x)   #do this before doing by groups   4/16/23
   if(!is.data.frame(group) && !is.list(group) && (length(group) < NROW(x))) group <- x[,group ,drop=FALSE]
		answer <- by(x,group,describe,type=type,...)
		answer <- fix.is.null(answer)
      	 class(answer) <- c("psych","describeBy")  #probably better not to make of class psych (at least not yet)
       	}

if (mat) { ncol <- length(answer[[1]])  #the more complicated case. How to reorder a list of data.frames
#the interesting problem is treating the case of multiple grouping variables. 
	n.var <- NROW(answer[[1]])
	n.col <- NCOL(answer[[1]])
	n.grouping <- length(dim(answer))  #this is the case of multiple grouping variables
	n.groups <- prod(dim(answer))
	names  <- names(answer[[1]])
	row.names <-attr(answer[[1]],"row.names")
	dim.names <- attr(answer,"dimnames")
  
	 mat.ans <- matrix(NA,ncol=ncol,nrow=n.var*n.groups) #NA or NaN
	 labels.ans <- matrix(NA,ncol = n.grouping+1,nrow= n.var*n.groups)  #CHANGE TO NA
	 colnames(labels.ans) <- c("item",paste("group",1:n.grouping,sep=""))
	 colnames(mat.ans) <- colnames(answer[[1]])
	 rn <- 1:(n.var*n.groups)
	 k <- 1
	 labels.ans[,1] <- seq(1,(n.var*n.groups))
	# for (grouping in 1:n.grouping)	   { labels.ans[,grouping+1] <- attr(answer,"dimnames")[[grouping]] }#no
	 group.scale <- cumprod(c(1,dim(answer)))

	 
	  for (var in 1:(n.var*n.groups)) {
	 		for (group in 1:n.grouping)  {
	     		groupi <- ((trunc((var-1)/group.scale[group]) ) %% dim(answer)[group] ) +1
	         
	          labels.ans[var,group+1] <- dim.names[[group]][[groupi]]}
	         }
	 
	 
	 k <- 1   
	 for (var  in 1:n.var) {
	  for (group in 1:n.groups) {
        
	  	rn[k] <- paste(row.names[var],group,sep="")
	    #mat.ans[k,1] <- group
	    for (stat in 1:n.col) {if(!is.null(answer[[group]][[stat]][var])) {
	   	 mat.ans[k,stat] <- round(answer[[group]][[stat]][var],digits)} else { mat.ans[k,stat] <- NA }
	   	  } 
	     k <- k + 1}
	   }
	   answer <- data.frame( labels.ans,mat.ans) 
	   rownames(answer) <- rn
        }
     
	  
	   #class(answer) <- c("psych","describe","list")
	  #answer$Call <- cl
return(answer)}


#added 11/05/20 to fix the problem of NULL values for certain combinations 
#doesn't work for 3 way or more classifications 1/5/20 (reported by Nicholas Stefaniak)
#fixed 1/5/20 

fix.is.null <- function(x ) {
fix <-  data.frame(vars=NA,n=NA,mean=NA,sd=NA,median=NA,trimmed=NA,mad=NA,min=NA,max=NA,range=NA,skew=NA,kurtosis=NA,se=NA)
fix <- char2numeric(fix)
n.obs <- NROW(x)
n.var <- NCOL(x)
n.cells <- length(x)
if(n.cells > n.obs * n.var) {
   for (cells in (1:n.cells))
   if(is.null(x[cells]))  {x[cells] <- fix
     }
   } else {
if(n.var > 1) {
for (i in 1:n.obs) {
   for (j in 1 :n.var) {
   if(is.null(x[[i,j]])) x[[i,j]] <- fix}
    }
    }}

return(x)
}