I am trying to obtain the Kruskal (1964) secondary least-squares monotonic
transformation of a rank variable given 4 categorical variables in order to
obtain optimal transformation for regression. The academic problem assigned
is to compare R, SPSS (Conjoint Analysis), and SAS' proc transreg in speed
and accuracy. Currently, SAS and SPSS are giving similar results, but R's
are quite different. There is something I am misunderstanding about
acepackage and/or isoMDS.
The data looks like this:
Brand Price Life Hazard Rank
1 Goodstone $69.99 60,000 Yes 3
2 Goodstone $69.99 70,000 No 2
...
7 Pirogi $69.99 50,000 No 7
8 Pirogi $69.99 70,000 No 1
9 Pirogi $74.99 50,000 Yes 8
The ace and avas functions transform the y values into very small values of
rank, like this:
$ty
[1] -1.3552125 -1.6732919 0.8859707
and hence the estimates are quite different.
The R-squared is .93 while SAS and SPSS give .99.
The isoMDS from MASS package gives weird results when i choose k=4.
Here is my acepackage code and isoMDS function:
X <- cbind(Brand, Price, Life, Hazard) # independent variables
Y <- Rank # response variable
cate <- as.vector(c(1,2,3,4)) # categorical variables(columns) in
X
mycon <- avas(x=X, y=Y, cat=cate)
mymatrix <- as.matrix((X)
row.names(mymatrix) <- Rank
Any help is well appreciated.
Thanks.
myc <- isoMDS(dist(mymatrix), k=?)
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
