R help - Jul 2017 - Quadratic function with interaction terms for the PLS fitting model?

poly(NIR, degree = 2) will work if NIR is a matrix, not a data.frame.
The degree argument apparently  *must* be explicitly named if NIR is
not a numeric vector. AFAICS, this is unclear or unstated in ?poly.


-- Bert

Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )


On Thu, Jul 13, 2017 at 10:15 AM, David Winsemius
<dwinsemius at comcast.net> wrote:
>
>> On Jul 12, 2017, at 6:58 PM, Ng, Kelvin Sai-cheong <kscng at
connect.hku.hk> wrote:
>>
>> Dear all,
>>
>> I am using the pls package of R to perform partial least square on a
set of
>> multivariate data.  Instead of fitting a linear model, I want to fit my
>> data with a quadratic function with interaction terms.  But I am not
sure
>> how.  I will use an example to illustrate my problem:
>>
>> Following the example in the PLS manual:
>> ## Read data
>> data(gasoline)
>> gasTrain <- gasoline[1:50,]
>> ## Perform PLS
>> gas1 <- plsr(octane ~ NIR, ncomp = 10, data = gasTrain, validation =
"LOO")
>>
>> where octane ~ NIR is the model that this example is fitting with.
>>
>> NIR is a collective of variables, i.e. NIR spectra consists of 401
diffuse
>> reflectance measurements from 900 to 1700 nm.
>>
>> Instead of fitting with predict.octane[i] = a[0] * NIR[0,i] + a[1] *
>> NIR[1,i] + ...
>> I want to fit the data with:
>> predict.octane[i] = a[0] * NIR[0,i] + a[1] * NIR[1,i] + ... +
>> b[0]*NIR[0,i]*NIR[0,i] + b[1] * NIR[0,i]*NIR[1,i] + ...
>>
>> i.e. quadratic with interaction terms.
>>
>> But I don't know how to formulate this.
>
> I did not see any terms in the model that I would have called interaction
terms. I'm seeing a desire for a polynomial function in NIR. For that
purpose, one might see if you get satisfactory results with:
>
> gas1 <- plsr(octane ~NIR + I(NIR^2), ncomp = 10, data = gasTrain,
validation = "LOO")
> gas1
>
> I first tried using poly(NIR, 2) on the RHS and it threw an error, which
raises concerns in my mind that this may not be a proper model. I have no
experience with the use of plsr or its underlying theory, so the fact that this
is not throwing an error is no guarantee of validity. Using this construction in
ordinary least squares regression has dangers with inferential statistics
because of the correlation of the linear and squared terms as well as likely
violation of homoscedasticity.
>
> --
> David.
>
>
>>
>> May I have some help please?
>>
>> Thanks,
>>
>> Kelvin
>>
>>       [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> 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.
>
> David Winsemius
> Alameda, CA, USA
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.

Bert,

The 'degree' argument follows the "..." argument in the
function declaration:

  poly(x, ..., degree = 1, coefs = NULL, raw = FALSE, simple = FALSE)

Generally, any arguments after the "..." must be explicitly named, but
as per the Details section of ?poly:

"Although formally degree should be named (as it follows ...), an unnamed
second argument of length 1 will be interpreted as the degree, such that poly(x,
3) can be used in formulas."

The issue of having to explicitly name arguments that follow the three dots has
come up over the years, but I cannot recall where that is documented in the
manuals.

Regards,

Marc


> On Jul 13, 2017, at 12:43 PM, Bert Gunter <bgunter.4567 at gmail.com>
wrote:
> 
> poly(NIR, degree = 2) will work if NIR is a matrix, not a data.frame.
> The degree argument apparently  *must* be explicitly named if NIR is
> not a numeric vector. AFAICS, this is unclear or unstated in ?poly.
> 
> 
> -- Bert
> 
> Bert Gunter
> 
> "The trouble with having an open mind is that people keep coming along
> and sticking things into it."
> -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip
)
> 
> 
> On Thu, Jul 13, 2017 at 10:15 AM, David Winsemius
> <dwinsemius at comcast.net> wrote:
>> 
>>> On Jul 12, 2017, at 6:58 PM, Ng, Kelvin Sai-cheong <kscng at
connect.hku.hk> wrote:
>>> 
>>> Dear all,
>>> 
>>> I am using the pls package of R to perform partial least square on
a set of
>>> multivariate data.  Instead of fitting a linear model, I want to
fit my
>>> data with a quadratic function with interaction terms.  But I am
not sure
>>> how.  I will use an example to illustrate my problem:
>>> 
>>> Following the example in the PLS manual:
>>> ## Read data
>>> data(gasoline)
>>> gasTrain <- gasoline[1:50,]
>>> ## Perform PLS
>>> gas1 <- plsr(octane ~ NIR, ncomp = 10, data = gasTrain,
validation = "LOO")
>>> 
>>> where octane ~ NIR is the model that this example is fitting with.
>>> 
>>> NIR is a collective of variables, i.e. NIR spectra consists of 401
diffuse
>>> reflectance measurements from 900 to 1700 nm.
>>> 
>>> Instead of fitting with predict.octane[i] = a[0] * NIR[0,i] + a[1]
*
>>> NIR[1,i] + ...
>>> I want to fit the data with:
>>> predict.octane[i] = a[0] * NIR[0,i] + a[1] * NIR[1,i] + ... +
>>> b[0]*NIR[0,i]*NIR[0,i] + b[1] * NIR[0,i]*NIR[1,i] + ...
>>> 
>>> i.e. quadratic with interaction terms.
>>> 
>>> But I don't know how to formulate this.
>> 
>> I did not see any terms in the model that I would have called
interaction terms. I'm seeing a desire for a polynomial function in NIR. For
that purpose, one might see if you get satisfactory results with:
>> 
>> gas1 <- plsr(octane ~NIR + I(NIR^2), ncomp = 10, data = gasTrain,
validation = "LOO")
>> gas1
>> 
>> I first tried using poly(NIR, 2) on the RHS and it threw an error,
which raises concerns in my mind that this may not be a proper model. I have no
experience with the use of plsr or its underlying theory, so the fact that this
is not throwing an error is no guarantee of validity. Using this construction in
ordinary least squares regression has dangers with inferential statistics
because of the correlation of the linear and squared terms as well as likely
violation of homoscedasticity.
>> 
>> --
>> David.
>> 
>> 
>>> 
>>> May I have some help please?
>>> 
>>> Thanks,
>>> 
>>> Kelvin
>>> 
>>>      [[alternative HTML version deleted]]
>>> 
>>> ______________________________________________
>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more,
see
>>> 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.
>> 
>> David Winsemius
>> Alameda, CA, USA
>> 
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> 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.
> 
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.

Marc:

1. I am aware of the need to explicitly name arguments after ... --
see the R Language definition where this can be inferred from the
argument matching rules.

2. I am aware of the stated exception for poly(). However:
> x1 <- runif(20)
> x2 <- runif(20)
> mx <- cbind(x1,x2)
> poly(mx,2)
Error in poly(dots[[i]], degree, raw = raw, simple = raw) :
  'degree' must be less than number of unique points
> poly(mx, degree = 2)
             1.0           2.0         0.1          1.1         0.2
 [1,] -0.2984843  0.0402593349 -0.07095761  0.021179734 -0.22909595
 [2,]  0.2512177  0.2172530896  0.29620999  0.074413206  0.14508422
 [3,]  0.2775652  0.3085750335 -0.13955410 -0.038735366 -0.13729529
 [4,] -0.4090782  0.4032189266 -0.14737858  0.060289370 -0.12358925
 [5,] -0.1631886 -0.2221937915 -0.26690975  0.043556631  0.16814432
 [6,]  0.1770952  0.0009863446  0.25380650  0.044947925  0.02737265
 [7,] -0.2108146 -0.1525957018  0.34023304 -0.071726094  0.28787441
 [8,]  0.2693983  0.2794576400  0.04697126  0.012653979 -0.26792015
 [9,]  0.2014353  0.0653896008 -0.37013148 -0.074557536  0.54445808
[10,] -0.1002967 -0.2761638672 -0.29389518  0.029476714  0.25539539
[11,]  0.1132090 -0.1372916959  0.21619808  0.024475573 -0.06074932
[12,] -0.1116108 -0.2696398425 -0.14592886  0.016287234 -0.12617869
[13,]  0.1792535  0.0064357827 -0.04948750 -0.008870809 -0.24736773
[14,] -0.1167216 -0.2662346206 -0.20209364  0.023588696 -0.00923419
[15,] -0.4258838  0.4700591049  0.08836730 -0.037634205 -0.24586894
[16,]  0.1047271 -0.1523001267 -0.21491954 -0.022507896  0.02225837
[17,] -0.1985753 -0.1728455549  0.32036901 -0.063617358  0.22084868
[18,]  0.1844006  0.0196368680  0.32321195  0.059600465  0.23017961
[19,]  0.1009775 -0.1586846110 -0.08282554 -0.008363512 -0.21685556
[20,]  0.1753745 -0.0033219134  0.09871464  0.017312033 -0.23746062
attr(,"degree")
[1] 1 2 1 2 2
attr(,"coefs")
attr(,"coefs")[[1]]
attr(,"coefs")[[1]]$alpha
[1] 0.5477073 0.4154115

attr(,"coefs")[[1]]$norm2
[1]  1.00000000 20.00000000  1.55009761  0.08065872

Cheers,
Bert


Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )


On Thu, Jul 13, 2017 at 11:17 AM, Marc Schwartz <marc_schwartz at me.com>
wrote:
> Bert,
>
> The 'degree' argument follows the "..." argument in the
function declaration:
>
>   poly(x, ..., degree = 1, coefs = NULL, raw = FALSE, simple = FALSE)
>
> Generally, any arguments after the "..." must be explicitly
named, but as per the Details section of ?poly:
>
> "Although formally degree should be named (as it follows ...), an
unnamed second argument of length 1 will be interpreted as the degree, such that
poly(x, 3) can be used in formulas."
>
> The issue of having to explicitly name arguments that follow the three dots
has come up over the years, but I cannot recall where that is documented in the
manuals.
>
> Regards,
>
> Marc
>
>
>
>> On Jul 13, 2017, at 12:43 PM, Bert Gunter <bgunter.4567 at
gmail.com> wrote:
>>
>> poly(NIR, degree = 2) will work if NIR is a matrix, not a data.frame.
>> The degree argument apparently  *must* be explicitly named if NIR is
>> not a numeric vector. AFAICS, this is unclear or unstated in ?poly.
>>
>>
>> -- Bert
>>
>> Bert Gunter
>>
>> "The trouble with having an open mind is that people keep coming
along
>> and sticking things into it."
>> -- Opus (aka Berkeley Breathed in his "Bloom County" comic
strip )
>>
>>
>> On Thu, Jul 13, 2017 at 10:15 AM, David Winsemius
>> <dwinsemius at comcast.net> wrote:
>>>
>>>> On Jul 12, 2017, at 6:58 PM, Ng, Kelvin Sai-cheong <kscng at
connect.hku.hk> wrote:
>>>>
>>>> Dear all,
>>>>
>>>> I am using the pls package of R to perform partial least square
on a set of
>>>> multivariate data.  Instead of fitting a linear model, I want
to fit my
>>>> data with a quadratic function with interaction terms.  But I
am not sure
>>>> how.  I will use an example to illustrate my problem:
>>>>
>>>> Following the example in the PLS manual:
>>>> ## Read data
>>>> data(gasoline)
>>>> gasTrain <- gasoline[1:50,]
>>>> ## Perform PLS
>>>> gas1 <- plsr(octane ~ NIR, ncomp = 10, data = gasTrain,
validation = "LOO")
>>>>
>>>> where octane ~ NIR is the model that this example is fitting
with.
>>>>
>>>> NIR is a collective of variables, i.e. NIR spectra consists of
401 diffuse
>>>> reflectance measurements from 900 to 1700 nm.
>>>>
>>>> Instead of fitting with predict.octane[i] = a[0] * NIR[0,i] +
a[1] *
>>>> NIR[1,i] + ...
>>>> I want to fit the data with:
>>>> predict.octane[i] = a[0] * NIR[0,i] + a[1] * NIR[1,i] + ... +
>>>> b[0]*NIR[0,i]*NIR[0,i] + b[1] * NIR[0,i]*NIR[1,i] + ...
>>>>
>>>> i.e. quadratic with interaction terms.
>>>>
>>>> But I don't know how to formulate this.
>>>
>>> I did not see any terms in the model that I would have called
interaction terms. I'm seeing a desire for a polynomial function in NIR. For
that purpose, one might see if you get satisfactory results with:
>>>
>>> gas1 <- plsr(octane ~NIR + I(NIR^2), ncomp = 10, data =
gasTrain, validation = "LOO")
>>> gas1
>>>
>>> I first tried using poly(NIR, 2) on the RHS and it threw an error,
which raises concerns in my mind that this may not be a proper model. I have no
experience with the use of plsr or its underlying theory, so the fact that this
is not throwing an error is no guarantee of validity. Using this construction in
ordinary least squares regression has dangers with inferential statistics
because of the correlation of the linear and squared terms as well as likely
violation of homoscedasticity.
>>>
>>> --
>>> David.
>>>
>>>
>>>>
>>>> May I have some help please?
>>>>
>>>> Thanks,
>>>>
>>>> Kelvin
>>>>
>>>>      [[alternative HTML version deleted]]
>>>>
>>>> ______________________________________________
>>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and
more, see
>>>> 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.
>>>
>>> David Winsemius
>>> Alameda, CA, USA
>>>
>>> ______________________________________________
>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more,
see
>>> 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.
>>
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> 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.
>

> On Jul 13, 2017, at 10:43 AM, Bert Gunter <bgunter.4567 at gmail.com>
wrote:
> 
> poly(NIR, degree = 2) will work if NIR is a matrix, not a data.frame.
> The degree argument apparently  *must* be explicitly named if NIR is
> not a numeric vector. AFAICS, this is unclear or unstated in ?poly.
I still get the same error with:

library(pld)
data(gasoline)
gasTrain <- gasoline[1:50,]
gas1 <- plsr(octane ~ poly(as.matrix(NIR), 2), ncomp = 10, data = gasTrain,
validation = "LOO")


Error in rep.int(rep.int(seq_len(nx), rep.int(rep.fac, nx)), orep) : 
  invalid 'times' value
> gas1 <- plsr(octane ~ poly(as.matrix(gasTrain$NIR), degree=2), ncomp =
10, data = gasTrain, validation = "CV")
Error in rep.int(rep.int(seq_len(nx), rep.int(rep.fac, nx)), orep) : 
  invalid 'times' value
> str(as.matrix(gasTrain$NIR))
 AsIs [1:50, 1:401] -0.0502 -0.0442 -0.0469 -0.0467 -0.0509 ...
 - attr(*, "dimnames")=List of 2
  ..$ : chr [1:50] "1" "2" "3" "4" ...
  ..$ : chr [1:401] "900 nm" "902 nm" "904 nm"
"906 nm" ...

So tried to strip the RHS down to a "simple" matrix
> gas1 <- plsr(octane ~ poly(matrix(gasTrain$NIR, nrow=nrow(gasTrain$NIR)
), degree=2), ncomp = 10, data = gasTrain, validation = "CV")
Error in rep.int(rep.int(seq_len(nx), rep.int(rep.fac, nx)), orep) : 
  invalid 'times' value

I guess it reflects my lack of understanding of poly (which parallels my lack of
understanding of PLS.)
-- 
David.
> 
> 
> -- Bert
> 
> Bert Gunter
> 
> "The trouble with having an open mind is that people keep coming along
> and sticking things into it."
> -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip
)
> 
> 
> On Thu, Jul 13, 2017 at 10:15 AM, David Winsemius
> <dwinsemius at comcast.net> wrote:
>> 
>>> On Jul 12, 2017, at 6:58 PM, Ng, Kelvin Sai-cheong <kscng at
connect.hku.hk> wrote:
>>> 
>>> Dear all,
>>> 
>>> I am using the pls package of R to perform partial least square on
a set of
>>> multivariate data.  Instead of fitting a linear model, I want to
fit my
>>> data with a quadratic function with interaction terms.  But I am
not sure
>>> how.  I will use an example to illustrate my problem:
>>> 
>>> Following the example in the PLS manual:
>>> ## Read data
>>> data(gasoline)
>>> gasTrain <- gasoline[1:50,]
>>> ## Perform PLS
>>> gas1 <- plsr(octane ~ NIR, ncomp = 10, data = gasTrain,
validation = "LOO")
>>> 
>>> where octane ~ NIR is the model that this example is fitting with.
>>> 
>>> NIR is a collective of variables, i.e. NIR spectra consists of 401
diffuse
>>> reflectance measurements from 900 to 1700 nm.
>>> 
>>> Instead of fitting with predict.octane[i] = a[0] * NIR[0,i] + a[1]
*
>>> NIR[1,i] + ...
>>> I want to fit the data with:
>>> predict.octane[i] = a[0] * NIR[0,i] + a[1] * NIR[1,i] + ... +
>>> b[0]*NIR[0,i]*NIR[0,i] + b[1] * NIR[0,i]*NIR[1,i] + ...
>>> 
>>> i.e. quadratic with interaction terms.
>>> 
>>> But I don't know how to formulate this.
>> 
>> I did not see any terms in the model that I would have called
interaction terms. I'm seeing a desire for a polynomial function in NIR. For
that purpose, one might see if you get satisfactory results with:
>> 
>> gas1 <- plsr(octane ~NIR + I(NIR^2), ncomp = 10, data = gasTrain,
validation = "LOO")
>> gas1
>> 
>> I first tried using poly(NIR, 2) on the RHS and it threw an error,
which raises concerns in my mind that this may not be a proper model. I have no
experience with the use of plsr or its underlying theory, so the fact that this
is not throwing an error is no guarantee of validity. Using this construction in
ordinary least squares regression has dangers with inferential statistics
because of the correlation of the linear and squared terms as well as likely
violation of homoscedasticity.
>> 
>> --
>> David.
>> 
>> 
>>> 
>>> May I have some help please?
>>> 
>>> Thanks,
>>> 
>>> Kelvin
>>> 
>>>      [[alternative HTML version deleted]]
>>> 
>>> ______________________________________________
>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more,
see
>>> 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.
>> 
>> David Winsemius
>> Alameda, CA, USA
>> 
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> 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.
David Winsemius
Alameda, CA, USA

David et.al.:

It's a problem with poly (or rather with how it is being misused)
> mx <- as.matrix(gasoline[1:50,"NIR"])
> str(mx)
 AsIs [1:50, 1:401] -0.0502 -0.0442 -0.0469 -0.0467 -0.0509 ...
 - attr(*, "dimnames")=List of 2
  ..$ : chr [1:50] "1" "2" "3" "4" ...
  ..$ : chr [1:401] "900 nm" "902 nm" "904 nm"
"906 nm" ...
> poly(mx[,1:5],2)  ## only 5 columns
Error in poly(dots[[i]], degree, raw = raw, simple = raw) :
  'degree' must be less than number of unique
points
> out <- poly(mx[,1:5], degree =2)
> dim(out)
[1] 50 20
## So this is same issue as before. But:
> out <- poly(mx[,1:30],degree = 2)
## 30 columns means 30*30 =900 2nd degree terms, but there are at most 50
## orthogonal vectors for the 50 -d space; ergo, poly() chokes rather
gracelessly, which is what you saw, with the following output:

rsession(2093,0x7fffe0c113c0) malloc: ***
mach_vm_map(size=823564528381952) failed (error code=3)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
rsession(2093,0x7fffe0c113c0) malloc: ***
mach_vm_map(size=823564528381952) failed (error code=3)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
Error: cannot allocate vector of size 767004.2 Gb

Cheers,
Bert


Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )


On Thu, Jul 13, 2017 at 4:36 PM, David Winsemius <dwinsemius at
comcast.net> wrote:
>
>> On Jul 13, 2017, at 10:43 AM, Bert Gunter <bgunter.4567 at
gmail.com> wrote:
>>
>> poly(NIR, degree = 2) will work if NIR is a matrix, not a data.frame.
>> The degree argument apparently  *must* be explicitly named if NIR is
>> not a numeric vector. AFAICS, this is unclear or unstated in ?poly.
>
> I still get the same error with:
>
> library(pld)
> data(gasoline)
> gasTrain <- gasoline[1:50,]
> gas1 <- plsr(octane ~ poly(as.matrix(NIR), 2), ncomp = 10, data =
gasTrain, validation = "LOO")
>
>
> Error in rep.int(rep.int(seq_len(nx), rep.int(rep.fac, nx)), orep) :
>   invalid 'times' value
>
>> gas1 <- plsr(octane ~ poly(as.matrix(gasTrain$NIR), degree=2), ncomp
= 10, data = gasTrain, validation = "CV")
> Error in rep.int(rep.int(seq_len(nx), rep.int(rep.fac, nx)), orep) :
>   invalid 'times' value
>
>> str(as.matrix(gasTrain$NIR))
>  AsIs [1:50, 1:401] -0.0502 -0.0442 -0.0469 -0.0467 -0.0509 ...
>  - attr(*, "dimnames")=List of 2
>   ..$ : chr [1:50] "1" "2" "3" "4"
...
>   ..$ : chr [1:401] "900 nm" "902 nm" "904
nm" "906 nm" ...
>
> So tried to strip the RHS down to a "simple" matrix
>
>> gas1 <- plsr(octane ~ poly(matrix(gasTrain$NIR,
nrow=nrow(gasTrain$NIR) ), degree=2), ncomp = 10, data = gasTrain, validation =
"CV")
> Error in rep.int(rep.int(seq_len(nx), rep.int(rep.fac, nx)), orep) :
>   invalid 'times' value
>
> I guess it reflects my lack of understanding of poly (which parallels my
lack of understanding of PLS.)
> --
> David.
>>
>>
>> -- Bert
>>
>> Bert Gunter
>>
>> "The trouble with having an open mind is that people keep coming
along
>> and sticking things into it."
>> -- Opus (aka Berkeley Breathed in his "Bloom County" comic
strip )
>>
>>
>> On Thu, Jul 13, 2017 at 10:15 AM, David Winsemius
>> <dwinsemius at comcast.net> wrote:
>>>
>>>> On Jul 12, 2017, at 6:58 PM, Ng, Kelvin Sai-cheong <kscng at
connect.hku.hk> wrote:
>>>>
>>>> Dear all,
>>>>
>>>> I am using the pls package of R to perform partial least square
on a set of
>>>> multivariate data.  Instead of fitting a linear model, I want
to fit my
>>>> data with a quadratic function with interaction terms.  But I
am not sure
>>>> how.  I will use an example to illustrate my problem:
>>>>
>>>> Following the example in the PLS manual:
>>>> ## Read data
>>>> data(gasoline)
>>>> gasTrain <- gasoline[1:50,]
>>>> ## Perform PLS
>>>> gas1 <- plsr(octane ~ NIR, ncomp = 10, data = gasTrain,
validation = "LOO")
>>>>
>>>> where octane ~ NIR is the model that this example is fitting
with.
>>>>
>>>> NIR is a collective of variables, i.e. NIR spectra consists of
401 diffuse
>>>> reflectance measurements from 900 to 1700 nm.
>>>>
>>>> Instead of fitting with predict.octane[i] = a[0] * NIR[0,i] +
a[1] *
>>>> NIR[1,i] + ...
>>>> I want to fit the data with:
>>>> predict.octane[i] = a[0] * NIR[0,i] + a[1] * NIR[1,i] + ... +
>>>> b[0]*NIR[0,i]*NIR[0,i] + b[1] * NIR[0,i]*NIR[1,i] + ...
>>>>
>>>> i.e. quadratic with interaction terms.
>>>>
>>>> But I don't know how to formulate this.
>>>
>>> I did not see any terms in the model that I would have called
interaction terms. I'm seeing a desire for a polynomial function in NIR. For
that purpose, one might see if you get satisfactory results with:
>>>
>>> gas1 <- plsr(octane ~NIR + I(NIR^2), ncomp = 10, data =
gasTrain, validation = "LOO")
>>> gas1
>>>
>>> I first tried using poly(NIR, 2) on the RHS and it threw an error,
which raises concerns in my mind that this may not be a proper model. I have no
experience with the use of plsr or its underlying theory, so the fact that this
is not throwing an error is no guarantee of validity. Using this construction in
ordinary least squares regression has dangers with inferential statistics
because of the correlation of the linear and squared terms as well as likely
violation of homoscedasticity.
>>>
>>> --
>>> David.
>>>
>>>
>>>>
>>>> May I have some help please?
>>>>
>>>> Thanks,
>>>>
>>>> Kelvin
>>>>
>>>>      [[alternative HTML version deleted]]
>>>>
>>>> ______________________________________________
>>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and
more, see
>>>> 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.
>>>
>>> David Winsemius
>>> Alameda, CA, USA
>>>
>>> ______________________________________________
>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more,
see
>>> 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.
>
> David Winsemius
> Alameda, CA, USA
>