RV: [R] dr ()

[De la Vega Góngora Jorge]

According to the algorithm, what is divided by the number of slices is the range of the response variable such that each slice has approximately the same number of  cases. Maybe the range of your response is very short or only takes a small number of values. I did a small simulation with normal data and what I got is the following:

 library(dr)
 X <- matrix(rnorm(3500*14),ncol=14) #Thinking you have 14 predictors. 
 y <- rnorm(3500,mean=8,sd=2) 
 msir <- dr(y~X,method="sir",nslices=8)
 summary(msir)

Call:
dr(formula = y ~ X, method = "sir", nslices = 8)

Method:
sir with 8 slices, n = 3500, using weights.

Slice Sizes:
438 438 438 438 437 437 437 437 

Eigenvectors:
          Dir1     Dir2      Dir3      Dir4
X1   0.2017030 -0.04039  0.192561  0.415955
X2  -0.0001632 -0.18160 -0.084247  0.544218
X3   0.1018999  0.04044 -0.214947 -0.497072
X4  -0.2591932  0.28825 -0.126928 -0.030583
X5  -0.3970003  0.27109 -0.194828 -0.041809
X6   0.4572450 -0.26988 -0.404183 -0.162565
X7  -0.0044665  0.51297  0.132349 -0.206052
X8  -0.0143563  0.31186  0.005307  0.337652
X9   0.4355999  0.37507  0.341414  0.006436
X10  0.0040755  0.03487 -0.441170  0.073009
X11 -0.0012413 -0.32851  0.092096 -0.102832
X12  0.3767253  0.30895 -0.412124  0.202124
X13 -0.3927512 -0.03363 -0.367138  0.149092
X14  0.1700059  0.16602 -0.224347  0.138282

               Dir1     Dir2     Dir3     Dir4
Eigenvalues 0.01195 0.006903 0.005031 0.003430
R^2(OLS|dr) 0.10372 0.641660 0.848887 0.878968

Asymp. Chi-square tests for dimension:
              Stat df p-value
0D vs >= 1D 106.69 98  0.2577
1D vs >= 2D  64.87 78  0.8560
2D vs >= 3D  40.71 60  0.9734
3D vs >= 4D  23.10 44  0.9960

Please let me know if you continue having problems on this.


Jorge de la Vega
 

-----Mensaje original-----
De: Jessica Higgs [mailto:jlh599 at psu.edu] 
Enviado el: Viernes, 22 de Abril de 2005 05:29 PM
Para: De la Vega Góngora Jorge
Asunto: RE: [R] dr ()


I have approximately 3500 observations. Even when I specify 8 slices, it 
does five with the first slice being significantly larger than the other 4.

At 05:16 PM 4/22/2005 -0500, you wrote:
>I think the method uses as default the number of slices such that each 
>slice has approximately the same number of data. How many observations 
>do you have?
>
>
>
>Jorge de la Vega
>
>
>-----Mensaje original-----
>De: r-help-bounces at stat.math.ethz.ch 
>[mailto:r-help-bounces at stat.math.ethz.ch] En nombre de Jessica Higgs 
>Enviado el: Viernes, 22 de Abril de 2005 01:48 PM
>Para: R-help at stat.math.ethz.ch
>Asunto: [R] dr ()
>
>
>Hi all--
>
>A quick question about the dr () function. I am using this function to
>reduce the dimensions of a data set I have that involves 14 predictor 
>variables and one predictant or response. The goal is to discover which 
>variables play the most important role in determining the response and, 
>thus, to reduce the variables. I would like to use the sliced inverse 
>regression method (SIR) within this function but each time I specify 8 
>slices, it only performs 5 slices. Any suggestions/thoughts?
>
>THanks,
>Jessica
>
>______________________________________________
>R-help at stat.math.ethz.ch mailing list 
>https://stat.ethz.ch/mailman/listinfo/r-help
>PLEASE do read the posting guide! 
>http://www.R-project.org/posting-guide.html