[R] Using Information from the Stats4 package in base envir

[lindsey at stat.fsu.edu]

Hi.  Thank you very much in advance for your help.

I have generated data from two simple linear models and used k-means  
clustering (stats4) to identify two clusters in the generated data.

Next, I would like to do simple linear regression for each separate  
cluster.  I can do this if I first use the cluster labels to define  
two separate data frames with the subset function.

However, I would ideally like to use the subset option in lm to  
identify the data for regression rather than creating separate data  
frames.  When I try to do it this way, I get the error, "Error in  
eval(expr, envir, enclos) : invalid 'envir' argument".  The code for  
this is given below.

If is is not possible to do this as an option within lm, is there  
another way rather than creating new data frames that are defined by a  
given cluster label?


Thanks again for your help.



library(stats4)

#Define the Models
#Two simple linear models:
#Model A: respA = a0 + a1x + e(a), e(a) ~N(0, sigma^2(a))
#Model B: respB = b0 + b1x + e(b), e(b) ~N(0, sigma^2(b))

a0= 0; a1 = 1.5; sigmaA=4; b0=50; b1=-2; sigmaB=4; n=250; min=0; max=50;

#Generate Data from the Models

	x1 = runif(n, min, max); #Explanatory variable
	eA = rnorm(n, 0, sigmaA);	eB = rnorm(n, 0, sigmaB);
	respA = a0 +a1*x1 + eA;		respB = b0 +b1*x1 + eB; #Responses


#K-Means Clustering

	resp1 = c(respA, respB);#Put response into single vector for k-means
	x12 = rep(x1,nummod);	#Put explanatory into a single vector
         data1 = data.frame(resp1, x12) #Data frame for kmeans
	cluster1 = kmeans(data1, 2, nstart=25)$cluster #Obtain cluster labels
	data1 = data.frame(data1 ,cluster1)#Cluster labels in third column
	data10 = subset(data1, cluster1==1)
	data11 = subset(data1, cluster1==2)

model10 = lm(resp1 ~ x12, data10)#It works using the subset data frame

model1 = lm(resp1 ~ x12, cluster1 == 1, data1); #Gives the following error

Error in eval(expr, envir, enclos) : invalid 'envir' argument