Logistic Regression on a 49G+.

[Joey C.]

I'm aware of another discussion involving logistic regression that
points to RAS 3.0:
http://www.hpcalc.org/details.php?id=1301

However, it is only for the 48 series. I tried using hpconv:
http://www.hpcalc.org/details.php?id=3799
To convert the 48 binary to a 49 one, but alas, it gave me numerous
"Address for entry xyz (abc) not found." errors.

Unfortunately, I don't know SysRPL, so I have no clue what I'm doing.
Would anyone be kind enough to help me out? Any suggestions for
converting this binary? Ideas of where to start?

Thank you.

--Joey C.
http://joeyjwc.x3fusion.com/

[John H Meyers]

On Mon, 04 Dec 2006 16:44:24 -0600, Joey C. wrote:

> I'm aware of another discussion involving logistic regression
> that points to RAS 3.0:
> http://www.hpcalc.org/details.php?id=1301

> However, it is only for the 48 series. I tried using hpconv:
> http://www.hpcalc.org/details.php?id=3799
> To convert the 48 binary to a 49 one, but alas, it gave me numerous
> "Address for entry xyz (abc) not found." errors.

Which is odd, since there's nothing "unsupported" in it.

> Unfortunately, I don't know SysRPL, so I have no clue what I'm doing.

> Any suggestions for converting this binary?

It's all *UserRPL*
(author MacDonald R. Phillips' favorite calculator language :),
except for one item 'PAD1' which is all "supported" SysRPL.

So what I'd do is:

o On HP48GX: Store dis-assembled text for 'PAD1' into 'PAD1'
o Recall entire directory and do STD 64 STWS \->STR
o Copy entire string to HP49 and do STR\-> (in Approx. mode)
o Store the directory.
o Compile 'PAD1' and store it back into 'PAD1'

Here's the source for 'PAD1'
(to save time with the first step above):

:: CK2NOLASTWD CK&DISPATCH0 # 31
:: DO>STR DUPLEN$ TWENTYTWO 4PICK LEN$ #-
2DUP#< ITE :: SWAP#- Blank$ SWAP&$ ;
:: 2DROP ; &$ ; ;

Don't have an HP48GX?
So use Emu48 or Debug4x -- it's cheaper than eBay :)

[r->] [OFF]

[John H Meyers]

Here are some things to watch out for
when attempting to simply transfer this UserRPL directory
from a 48 series calc to a 49 series calc:

> logistic regression, RAS 3.0:
> http://www.hpcalc.org/details.php?id=1301

There is one very small SysRPL program 'PAD1' (as mentioned earlier),
which needs to be converted to source and recompiled after transfer.

There is a variable named 'ANS' (referenced in two other places);
however, ANS is a *command* in 49 series, so you'll have to change
that name in all three places.

It also contains *unsupported* #353ABh [HP48] SYSEVAL in two places;
thus far, #2F110h SYSEVAL (or #2B7001h FLASHEVAL) does the same
in 49G[+]/50G/48Gii, but these still aren't "supported"
(use at your own risk); or you can use LNAME NIP AXL instead
(works with most expressions, not at all with programs).

This offers a few examples of why even "UserRPL" transfer
needs close attention.

[r->] [OFF]

[John H Meyers]

Original file for HP48GX:

> logistic regression, RAS 3.0:
> http://www.hpcalc.org/details.php?id=1301

Here's a downloadable Zip file for 49G[+]/50G/48Gii
with source and binary (will remain posted for 7 days):

File Name: ras30B.zip
Size: 43.29 KB

http://www.filefactory.com/file/775cc5/

[r->] [OFF]

[werty]

48 is emulated on 49 !!! Not a very good job of emulation !

not even a good job of Saturn emulation .

HP is history , they are not selling pocket calc's .

people are waiting for the PDA with 120 GB HDD to
destroy all PC's , notebooks , pocket calcs ...everything .

A new free OpSys and the 120GB HDD
wont be from Intel, M$ , HP , ....

it will be from hobbyists ! free ....

[John H Meyers]

On Tue, 05 Dec 2006 05:25:46 -0600, I wrote:

> Here's a downloadable Zip file for 49G[+]/50G/48Gii
> with source and binary (will remain posted for 7 days)

Source file was okay, compiled binary had problem
(because I had a *variable* named 'CHOOSE' in my calc,
for a full-screen version of CHOOSE, so every CHOOSE
got compiled into a reference to that variable,
rather than a reference to the built-in command).

Temporary solution:

The entire single directory object is pure UserRPL;
just EDIT (then press ENTER, making no changes),
purge the original whole directory and store the new result
(should be fixed in update sent to Eric for www.hpcalc.org)

Any other problem is probably the same as the original for 48GX,
inasmuch as the only change was to replace a simple SysRPL
blank-padding program (called only once) with a UserRPL equivalent
(the utility of which is now to make the entire directory
editable as one object, just as above :)

Now excuse me while I regress back to my own mean self...

[r->] [OFF]