what happened to start=(row, col) in `symbolicMatrix()`?

[friendly]

Before the flurry of generalizing w/ Eqn() et al, there was an arg start = c(row=1, col=1) that could be be used to start the indexing of rows/cols at 0 insymbolicMatrix()`. Just found myself in need of this, and realized it disappeared in the dust. How much work would it take to put it back?

[john-d-fox]

Hi,

On 2024-08-09 2:54 p.m., Michael Friendly wrote:

Caution: External email.

Before the flurry of generalizing w/ Eqn() et al, there was an arg | start = c(row=1, col=1) that could be be used to start the indexing of rows/cols at 0 in |symbolicMatrix()`. Just found myself in need of this, and realized it disappeared in the dust. How much work would it take to put it back?

I don't recall this. Can you find it in the GitHub history?

John

— Reply to this email directly, view it on GitHub <https://github.com/ friendly/matlib/issues/53>, or unsubscribe <https://github.com/ notifications/unsubscribe-auth/ ADLSAQRC3C5TPVU3GBCZ2GLZQUF7BAVCNFSM6AAAAABMI7COK6VHI2DSMVQWIX3LMV43ASLTON2WKOZSGQ2TQNBZGQ3TENQ>. You are receiving this because you are subscribed to this thread.Message ID: @.***>

[friendly]

Ugh! I tried git blame R/symbolicMatrix.R and the blame button on GitHub, but can't find this. The history goes back to Aug 2. I had an older copy in dev/, but deleted it when I cleaned up.

[john-d-fox]

Hi,

Unless I'm completely losing it, I never explicitly enabled 0-based indexing in symbolicMatrix(). Maybe it worked anyway, but I don't see how.

OTOH, I don't think that it would be hard to implement in the latest approach I'm taking to the function. Would you like me to try? If so, I think that I'd prefer to have a logical argument, say zero.based, which defaults to FALSE and applies to both rows and columns. Any reason not to use that rather than a 2-element numeric start argument?

Also, I can't say that I've ever enjoyed working with programming languages that use 0-based indexing and wonder why it would be useful for something like symbolicMatrix().

Best, John

On 2024-08-09 3:41 p.m., Michael Friendly wrote:

Caution: External email.

Ugh! I tried |git blame R/symbolicMatrix.R| and the blame button on GitHub, but can't find this. The history goes back to Aug 2. I had an older copy in |dev/|, but deleted it when I cleaned up.

— Reply to this email directly, view it on GitHub <https://github.com/ friendly/matlib/issues/53#issuecomment-2278627614>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ ADLSAQSCJTMCDBQOKRZPEYTZQULQBAVCNFSM6AAAAABMI7COK6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENZYGYZDONRRGQ>. You are receiving this because you commented.Message ID: <friendly/ @.***>

[john-d-fox]

Hi again,

Since you didn't object, I added a zero.based arg to symbolicMatrix() in dev/symbolicMatrix.R. It seems to work fine; for example,

symbolicMatrix(zero.based=TRUE) symbolicMatrix(nrow=3, ncol=4, zero.based=TRUE) symbolicMatrix(ncol=4, zero.based=TRUE) symbolicMatrix(nrow=3, zero.based=TRUE) symbolicMatrix(nrow="n", ncol="n", diag=TRUE, zero.based=TRUE) symbolicMatrix(nrow=3, ncol=3, diag=TRUE, zero.based=TRUE)

Everything else should work as before but of course more checking is eventually in order.

Best, John

On 2024-08-09 3:41 p.m., Michael Friendly wrote:

Caution: External email.

Ugh! I tried |git blame R/symbolicMatrix.R| and the blame button on GitHub, but can't find this. The history goes back to Aug 2. I had an older copy in |dev/|, but deleted it when I cleaned up.

— Reply to this email directly, view it on GitHub <https://github.com/ friendly/matlib/issues/53#issuecomment-2278627614>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ ADLSAQSCJTMCDBQOKRZPEYTZQULQBAVCNFSM6AAAAABMI7COK6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENZYGYZDONRRGQ>. You are receiving this because you commented.Message ID: <friendly/ @.***>

[friendly]

sorry for not responding earlier... I'm entering packing mode for our holiday, starting Monday.

It must have been @philchalmers who implemented the possibility of zero-based indexing, The use case for separate row/col bases is the $\mathbf{B}_{(q+1) \times p}$ matrix in a multivariate regression problem, where the first row $\mathbf{b}_0^\top$ is for the intercept but the columns for the responses are numbered 1, ... ,p.

$$ \sizedmat{B}{(q+1) \times p} = \begin{pmatrix} b{0}^\top \ b{1}^\top \ \vdots \ b{i}^\top \ \vdots \ b{q}^\top \ \end{pmatrix} $$

[philchalmers]

Not I, but the commit to add start was here: https://github.com/friendly/matlib/commit/8c19247da3d5e496895f2232952f525201136f68

If you need access to this file create a branch at this exact commit and copy the file out:

git checkout 8c19247da3d5e496895f2232952f525201136f68
cp R/symb-matrix.R some/u/want/symb-matrix.R
git checkout master

[john-d-fox]

Hi,

OK, it wasn't hard to make zero.based a two-element logical vector for rows and columns, so now

symbolicMatrix(zero.based=c(TRUE, TRUE)) symbolicMatrix(nrow=3, ncol=4, zero.based=c(TRUE, TRUE)) symbolicMatrix(ncol=4, zero.based=c(TRUE, TRUE)) symbolicMatrix(nrow=3, zero.based=c(TRUE, TRUE)) symbolicMatrix(nrow="n", ncol="n", diag=TRUE, zero.based=c(TRUE, TRUE)) symbolicMatrix(nrow=3, ncol=3, diag=TRUE, zero.based=c(TRUE, TRUE))

symbolicMatrix(zero.based=c(TRUE, FALSE)) symbolicMatrix(zero.based=c(FALSE, TRUE))

symbolicMatrix("\mathbf{B}", nrow="q", ncol="p",

• zero.based=c(TRUE, FALSE)) \begin{pmatrix} \mathbf{B}{0,1} & \mathbf{B}{0,2} & \cdots & \mathbf{B}{0,p} \ \mathbf{B}{1,1} & \mathbf{B}{1,2} & \cdots & \mathbf{B}{1,p} \ \vdots & \vdots & & \vdots \ \mathbf{B}{q - 1,1} & \mathbf{B}{q - 1,2} & \cdots & \mathbf{B}_{q
• 1,p} \ \end{pmatrix}

You'll notice that the last doesn't really produce what you want. Expecting to get 0, 1, ..., q if nrow="q" isn't really 0-based indexing in my understanding of the term, because there would then be q + 1 rows.

The following also isn't satisfactory, because the function doesn't do arithmetic if nrow is character:

symbolicMatrix("\mathbf{B}", nrow="q + 1", ncol="p",

• zero.based=c(TRUE, FALSE)) \begin{pmatrix} \mathbf{B}{0,1} & \mathbf{B}{0,2} & \cdots & \mathbf{B}{0,p} \ \mathbf{B}{1,1} & \mathbf{B}{1,2} & \cdots & \mathbf{B}{1,p} \ \vdots & \vdots & & \vdots \ \mathbf{B}{q + 1 - 1,1} & \mathbf{B}{q + 1 - 1,2} & \cdots & \mathbf{B}_{q + 1 - 1,p} \ \end{pmatrix}

Best, John

On 2024-08-09 5:20 p.m., Michael Friendly wrote:

Caution: External email.

sorry for not responding earlier... I'm entering packing mode for our holiday, starting Monday.

It must have been @philchalmers https://github.com/philchalmers who implemented the possibility of zero-based indexing, The use case for separate row/col bases is the $\mathbf{B}_{(q+1) \times p}$ matrix in a multivariate regression problem, where the first row $\mathbf{b}_0^\top$ is for the intercept but the columns for the responses are numbered 1, ... ,p.

$$ \sizedmat{B}{(q+1) \times p} = \begin{pmatrix} b{0}^\top \ b{1} ^\top \ \vdots \ b{i}^\top \ \vdots \ b{q}^\top \ \end{pmatrix} $$

— Reply to this email directly, view it on GitHub <https://github.com/ friendly/matlib/issues/53#issuecomment-2278770762>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ ADLSAQX5MOQ2A77H7PCZXULZQUXABAVCNFSM6AAAAABMI7COK6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENZYG43TANZWGI>. You are receiving this because you commented.Message ID: <friendly/ @.***>

[john-d-fox]

Hi,

This looks like a completely different function, which Michael wrote.

Best, John

On 2024-08-09 5:49 p.m., Phil Chalmers wrote:

Caution: External email.

Not I, but the commit to add |start| was here: 8c19247 <https:// github.com/friendly/matlib/commit/8c19247da3d5e496895f2232952f525201136f68>

If you need access to this file create a branch at this exact commit and copy the file out:

|git checkout 8c19247da3d5e496895f2232952f525201136f68 cp R/symb- matrix.R some/u/want/symb-matrix.R git checkout master |

— Reply to this email directly, view it on GitHub <https://github.com/ friendly/matlib/issues/53#issuecomment-2278797288>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ ADLSAQTOWD6Q6BJ6P5KT4YTZQU2O5AVCNFSM6AAAAABMI7COK6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENZYG44TOMRYHA>. You are receiving this because you commented.Message ID: <friendly/ @.***>

[philchalmers]

Indeed, it was before the port to the new symbolicMatrix occurred. That's why there was no history of it via git blame.

[john-d-fox]

Hi Michael,

After some messing around, and at the expense of proliferating arguments, I think that I have a solution that can produce what you want:

symbolicMatrix("\mathbf{B}", nrow="q", ncol="p",

• zero.based=c(TRUE, FALSE),
• end.at.n.minus.1=c(FALSE, TRUE)) \begin{pmatrix} \mathbf{B}{0,1} & \mathbf{B}{0,2} & \cdots & \mathbf{B}{0,p} \ \mathbf{B}{1,1} & \mathbf{B}{1,2} & \cdots & \mathbf{B}{1,p} \ \vdots & \vdots & & \vdots \ \mathbf{B}{q,1} & \mathbf{B}{q,2} & \cdots & \mathbf{B}_{q,p} \ \end{pmatrix}

The argument zero.based has entries for rows and columns, and whether 0-based indices run to nrow/ncol or nrow-1/ncol-1 is controlled by the argument end.at.n.minus.1.

Let me know what you think, John

On 2024-08-09 5:20 p.m., Michael Friendly wrote:

Caution: External email.

sorry for not responding earlier... I'm entering packing mode for our holiday, starting Monday.

It must have been @philchalmers https://github.com/philchalmers who implemented the possibility of zero-based indexing, The use case for separate row/col bases is the $\mathbf{B}_{(q+1) \times p}$ matrix in a multivariate regression problem, where the first row $\mathbf{b}_0^\top$ is for the intercept but the columns for the responses are numbered 1, ... ,p.

$$ \sizedmat{B}{(q+1) \times p} = \begin{pmatrix} b{0}^\top \ b{1} ^\top \ \vdots \ b{i}^\top \ \vdots \ b{q}^\top \ \end{pmatrix} $$

— Reply to this email directly, view it on GitHub <https://github.com/ friendly/matlib/issues/53#issuecomment-2278770762>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ ADLSAQX5MOQ2A77H7PCZXULZQUXABAVCNFSM6AAAAABMI7COK6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENZYG43TANZWGI>. You are receiving this because you commented.Message ID: <friendly/ @.***>

[philchalmers]

@friendly has this been resolved?

[friendly]

Yes, we now have zero.based and end.at args. I'll close this now,