R help - Dec 2011 - Very strange function() behaviour.

Hi, R newb here. I've coded a function that inputs N dimensional array(s)
[or
class=numeric if it's dim=1] of coefficients and tstats, where
dim(coef_matrix)=dim(tstat_matrix), it will then output a same dimension
matrix of coefficients pasted to tstats in brackets pasted to significance
stars.

If I go straight to the code INSIDE the function, it works 100% as it's
supposed to. The output matrix has the correct dimensions. However, if I
actually use the function, it'll do something like this.

fun.starpaste(3D matrix tstats,3D matrix coefficients) ::: CORRECT OUTPUT, 3
dimensional array.
fun.starpaste(1D numeric tstats, 1D numeric coefficients) ::: WRONG output,
output is a 3 dimensional array. And it'll keep returning a 3 dimensional
array (it'll be the correct 1 dimensional vector repeated numerous times to
fill 3 dimensions).

Then I'll copy and paste the function code again and again into the R
console, and it'll reset the slate:

fun.starpaste(1D numeric, 1D numeric) ::: CORRECT output, 1 dimensional
vector.
fun.starpaste(2D or 3D matrix, 2D or 3D matrix) ::: WRONG output, gives me a
massive 1 dimensional vector.

What the heck?? The function code is below.


#----------------------------------------------------------------------------------------------

             fun.starpaste <-
function(t,co=NULL,tails=2,sig=c(0.9,0.95,0.99))   #handles the pasting of
#coef(tstat)significance_stars
  
             {

if(length(dim(t))>1)t3<-t else fake1<-1
t<-matrix(t)

if(is.null(co)==TRUE)fake<-1 else co<-matrix(co)

if(exists(as.character(substitute(fake1)))==FALSE)fake<-1 else t3<-t			 
			 
tstat<-as.numeric(t);tstat[is.na(tstat)]<-0

sig.l<-length(sig)
stars<-rawToChar(charToRaw("*********************************************"),multiple=T)


MISC1<-0
  for(x in 1:length(sig)){
     if(tails==1)cutoff<-qt(sig[x],Inf) else
if(tails==2)cutoff<-qt(sig[x]+0.5*(1-sig[x]),Inf) else fake<-1
     MISC1<-cbind(MISC1,cutoff)
  }
cutoff<-MISC1[-1]


  for(i in 1:length(tstat)){ 
     if(abs(as.numeric(tstat[i]))<=cutoff[1])tstat[i]<-c("")
else fake<-1
  }
  for(i in 1:length(tstat)){ 
  for(x in 1:(sig.l-1)){
     if(sum(apply(matrix(1:(sig.l+1),nrow=1),c(1,2),function(N){
if(tstat[i]==cbind(c(""),c("*"),c("**"),c("***"),c("****"),c("*****"),c("******"))[N])1
else 0  } ))==0 && abs(as.numeric(tstat[i]))>cutoff[x] &&
abs(as.numeric(tstat[i]))<cutoff[x+1])            
tstat[i]&lt;-paste(stars[1:x],collapse=&quot;&quot;,sep=&quot;&quot;)
else
fake&lt;-1
	 if(sum(apply(matrix(1:(sig.l+1),nrow=1),c(1,2),function(N){
if(tstat[i]==cbind(c(&quot;&quot;),c(&quot;*&quot;),c(&quot;**&quot;),c(&quot;***&quot;),c(&quot;****&quot;),c(&quot;*****&quot;),c(&quot;******&quot;))[N])1
else 0  } ))==0 &amp;&amp; abs(as.numeric(tstat[i]))>=cutoff[sig.l])
tstat[i]<-paste(stars[1:sig.l],collapse="",sep="") else
fake<-1
  }}
options(warn=-1)
a<-as.numeric(tstat);b<-a[!is.na(a)]
if(sum(b)>0){a<-cbind(1:length(tstat),a);a<-a[!is.na(a[,2]),1];tstat[a]<-c("ERROR")}
else fake<-1  

data2<-array(tstat,dim=dim(t))
dimnames(data2)<-dimnames(t)


t2<-format(round(t,tround),nsmall=tround,trim=T)
 if(is.null(co)==FALSE){
co<-format(round(co,coefround),nsmall=coefround,trim=T)
t<-paste(co,c("("),t2,c(")"),data2,sep="")
t<-array(t,dim=c(length(t),1))
dimnames(t)<-dimnames(data2)
 } else {t<-data2}
if(length(dim(t3))>1 &&
length(dim(t3)[dim(t3)==1])==0)t<-array(t,dim=dim(t3)) else fake<-1
return(t)

		     }

-----
----

Isaac
Research Assistant
Quantitative Finance Faculty, UTS
--
View this message in context:
http://r.789695.n4.nabble.com/Very-strange-function-behaviour-tp4247729p4247729.html
Sent from the R help mailing list archive at Nabble.com.

Okay it's working perfectly now. I restarted R and it worked on my first 5
goes.

Can anybody shed light on how this kind of thing can happen?

-----
----

Isaac
Research Assistant
Quantitative Finance Faculty, UTS
--
View this message in context:
http://r.789695.n4.nabble.com/Very-strange-function-behaviour-tp4247729p4247740.html
Sent from the R help mailing list archive at Nabble.com.

Sorry if there's an easy answer to this problem, but here goes..
*
INTRODUCTION & CONTEXT*

I'm creating a function where the number of entries in the lm(y~.......)
varies. i.e. depending on the function input I want lm(y~x1+x2+x3),
sometimes I'll want lm(y~x1) only, et cetera.

I've completed this with paste(..,sep="") and as.formula().

*The problem*

Many of these possible entries [i.e. the x1, x2, x3] need their own lapply
for the lm() to work, as they're all lists or matrices with many dimensions
[I'm iterating along the dim of each X, which is why I need separate
lapply() for each]. Except, I can't do as.formula("..") with
lapply(),
therefore I'm stuffed!

This is the problem area:

#--------------------------------------------------
rl<-list()
for(Q in 1:ncol(y)){
rl[[colnames(y)[Q]]]<-
lapply(1:length(x1),function(N){lm(y[,Q]~x1[[N]])}
}
#-------------------------------------------------

This is the correct code for a case of 1 independent variable. But if the
preceding code wants to put in x2 or x3 (which it can do fine), how do I get
more lapplys to automatically layer on top of the above code, responsive to
the number of x? [[Note I can't layer loops on top of it... as I'm
shoving
the lm() into a list(), lapply() puts the lm() in the correct dimensions of
the list() but for() loops do not.]]

##P.S. just consider y to be:
y<-matrix(rnorm(100),ncol=50,nrow=100)
##x1, x2, x3, etc are either 1 or 2 dimensional lists... Just know that
there needs to be some lapply()- or loop-esque "thingy" that deals
with each
one, otherwise it won't work.


Thank you everyone.


-----
----

Isaac
Research Assistant
Quantitative Finance Faculty, UTS
--
View this message in context:
http://r.789695.n4.nabble.com/Very-strange-function-behaviour-tp4252877p4252877.html
Sent from the R help mailing list archive at Nabble.com.