Hello all-
I am a relatively new user of R and am working through a graduate course
in
Statistics that uses Minitab, SAS and some Matlab. I like using R but
am
having some trouble lining up the output from lda() to that of the other
programs'
results. The dataset below is a modified set of wine data from the
Pinot Noir
data set as an illustration of the 2 group LDA scenario.
Mo Ba Region
1 0.058 0.225 3
2 0.071 0.105 3
3 0.147 0.301 2
4 0.116 0.166 3
5 0.166 0.132 3
6 0.261 0.078 3
7 0.191 0.085 3
8 0.009 0.072 3
9 0.027 0.094 3
10 0.030 0.349 2
11 0.268 0.099 3
12 0.245 0.071 3
13 0.161 0.181 2
14 0.146 0.328 2
15 0.155 0.081 3
16 0.126 0.299 2
17 0.211 0.206 2
18 0.129 0.281 2
19 0.166 0.292 2
20 0.199 0.292 2
21 0.208 0.087 3
There are various displays from SAS, Minitab and Matlab's Statistics
Toolbox, but
they all agree with the final linear discriminant function: 0 = -19.51 +
21.47*Mo + 84.08*Ba.
When I run the following code in R:
> library(MASS)
> wine.lda <- lda(Region ~ Mo + Ba, data = wine, prior = c(1,1)/2)
> wine.lda
Call:
lda(Region ~ Mo + Ba, data = wine, prior = c(1, 1)/2)
Prior probabilities of groups:
2 3
0.5 0.5
Group means:
Mo Ba
2 0.1461111 0.2810000
3 0.1479167 0.1079167
Coefficients of linear discriminants:
LD1
Mo -5.636024
Ba -22.069187
I am having trouble going from the cofficients derived from the
spherical within group covariance to the
function I am able to obtain in the other programs.
The rest of the problem set up for all programs are: Region = group
assignment, prior = equal priors and
I do not do any data pretreatment prior to analysis.
Any help would be great.
Thanks,
Matt
[[alternative HTML version deleted]]
______________________________________________
[email protected] 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.
