Message-Id: <m0roKkG-00007ZC@xiron.pc.helsinki.fi>
Date:         Mon, 13 Mar 1995 19:49:30 EST
Reply-To:     jorge@PHYAST.PITT.EDU
Sender:       Lojban list <LOJBAN%CUVMB.bitnet@FINHUTC.hut.fi>
From:         jorge@PHYAST.PITT.EDU
Subject:      Re: Numbers
To:           Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Content-Length: 3461
Lines: 71

la kris cusku di'e (to tebe'i mi po'o i ku'i tecu'u le liste toi)

> I imagine that John's method won't actually be intrerpreted as 20ji'i02 -->
> 2000-2099 with mean 2002, since, as you say, it would be a weird coincidence
> if you just happened to have a range that fell on those particular numbers.
> I was guessing that pragmatically it would be interpreted as "2002, give or
> take about 50" (even if it were 20ji'i01).  So in practice, if I guess right
> about the pragmatics, it lets you specify the mean exactly, and the range
> within a power of ten.

That makes more sense, but then let's not bring in non-symmetric ranges.

> Your system doesn't seem to allow that.  The range of accuracy is a function
> of where the zeroes happen to fall; I can't distinguish between 2000
> plus-or-minus 5 and 2000 plus-or-minus 500, since you only consider the
> number to have one digit of accuracy.  I guess if I wanted plus-or-minus 5 I
> could say 2001ji'i (putting the one in to flag the last significant digit).

Yes, that's what I proposed in one of my posts. Since it is an approximate
number anyway it doesn't really matter if you change the 0 to a 1 for the
last significant (or rather the first non-significant) digit. I don't think
{ji'i} is enough for precise notation anyway, it is just for everyday use.

> Here's another proposal; how about if the first number is the mean and the
> second number is 1/2 the expected range, and it's assumed that the range is
> symmetric.  That way you get exactly the range you want and the mean you
> want; if you leave off the second argument it defaults to Jorge's rule about
> significant digits.

That could be a possibility, but I prefer to give an approximate range
with the two numbers. Maybe it is malrarna, but I find it the easiest way
to give approximate numbers.

Second message:

> OOPS! for 2 reasons:
>
> 1) I sent that last message (about numbers) to you rather than to the Lojban
> list (I wish the list server would send us stuff with the jboste as the
> return address, rather than the original sender -- I'm always forgetting to
> change it back)  So noone but you has seen it.  Which is just as well
> because la'edi'e

I now sent it to the whole list, I hope it's ok with you.

> 2) I forgot that your original message actually put a range around ji'i, not
> just using one side of it as I misread and stated here.  Given that, I have
> much less objection to it.

I propose both uses. A single number next to ji'i says that the number is
approximate (with the range of uncertainty given by the last non-zero digit)
If you give two numbers with ji'i in between, then it means that it is a
number approximately in that range.

I find it very useful to translate things like "two or three people",
"twenty or thirty cows", "he's thirty five, thirty six years old".

It could be done by specifying a mean plus a range, but that involves
more mental calculation. With my method you simply give two numbers that
more or less "fit". There's no need to calculate what would be the range.

With the interposed ji'i it is even worse. You have to decide where
to split the number so that you get the right range. I think that
having to make all those calculations on the fly will make the use
of ji'i very difficult, if not impossible.

I can't think of any occasion where one would need to specify the mean
with an accuracy beyond that of the range, which is what the interposed
ji'i allows.

Jorge