negative log scale

[Ingo Wardinski]

Hi,
I would like to render the attached data in a logarithmic style, but
as there are some negative values, the plot looks sketchy. I'm aware
of the fact that log(x) is only defined for x>0. So, what I really
would like to know is there a way to declare a logarithmic x-axis with
negative tics and then plot the attached data using

pl 'attached_data' us 1:($2<0?abs($2)*(-1.0):$2) w l

many thanx in advance
ingo

data
1 890.976
2 1.47644E-03
3 -22132.1
4 -19532.0
5 22132.0
6 19532.0
7 -0.125568
8 -5.21421E-02
9 -3.20301E-02
10 6.71634E-02
11 0.142713
12 -0.255993
13 0.448785
14 0.346983
15 0.578917
16 -0.339599
17 -9.12664
18 -4.85972
19 -9.82783
20 140.229
21 794.451
22 193.238
23 -825.706
24 -2202.26

--
ingo wardinski GeoForschungsZentrum Potsdam, in...@gfz-potsdam.de
**********************************************************************

"I did my graduate work on Mars." E. Theilig Galileo Project manager

[Petrik Galvosas]

Hi,

> So, what I really would like to know is there a way to declare a
> logarithmic x-axis

Do you really mean that? I suppose you mean y-axis.

> pl 'attached_data' us 1:($2<0?abs($2)*(-1.0):$2) w l

As far as I can see this is identical to
pl 'attached_data' w l
since you attempted to plot the negative value of the absolute value in
case the value is negative. Which is a negative value.

You might get somehow what you want if you plot

pl 'attached_data' 1:($2>0?log($2):-log(-$2)) w l

but be aware of the "cruelty" you do to math :-).
Bye Petrik

[Aidan Karley]

I'm trying to work out what you could be wanting to do this for, which
would prevent you from simply moving your data by a constant amount to
make it all positive. That would get rid of your apparent problem
without "folding" the data about your zero line.
Incidentally, since you test for parameter 2 being less than 0, then
taking the absolute of the value and multiplying it by minus one runs
you straight back into the "log(x) is only defined for x>0" problem.
Is the precision of your data really sufficient that you're
needing to plot to 13 significant figures. Because trying to put that
onto a log scale is going to mean that your (transformed) zero line is
still going to be at one millionth of the scale of the top of the axis
(and a similar factor from the scaling at the other end of the axis).

I'm really puzzled by what you're trying to do. Is it economics
data or something, of physics data. I find the presence of this series

> 3 -22132.1
> 4 -19532.0
> 5 22132.0
> 6 19532.0
>

suggestive that this is artificial data, not from any physical
process I can readily envisage.

--
Aidan Karley,
Aberdeen, Scotland,
Location: 57°10'11" N, 02°08'43" W (sub-tropical Aberdeen), 0.021233

[Ingo Wardinski]

Thanx to Petrik and Aidan for getting myself sorted.

pl 'attached_data' 1:($2>0?log($2):-log(-$2)) w l

does the job. The data I posted are a kind of spectrum, to answer
Aidan's question.
sorry for my low brain on saturday
cheers
ingo

--
ingo wardinski GeoForschungsZentrum Potsdam, in...@gfz-potsdam.de

**********************************************************************
echo '[dO%O+38%O+PO/d0<0]Fi22os0CC4BA64E418CE7l0xAP'|dc

[Aidan Karley]

In article <ynlllfx...@gfz-potsdam.de>, Ingo Wardinski wrote:
> The data I posted are a kind of spectrum, to answer
> Aidan's question.
> sorry for my low brain on saturday
> cheers
> ingo
> --
> ingo wardinski GeoForschungsZentrum Potsdam, in...@gfz-potsdam.de
>

Oh, I should have noticed the "gfz-potsdam" in that address.
Crazy geophysicists, should have stuck to hitting the rocks with
a hammer. Or a ton of explosives and a drilling bit <G>.
I guess I'll have to go and look at the KTB website now, to see
if there's been anything since my last visit.

[Ingo Wardinski]

>>>>> "AK" == Aidan Karley <ai...@abuse.demon.co.uk.invalidated> writes:

AK> Oh, I should have noticed the "gfz-potsdam" in that address.
AK> Crazy geophysicists, should have stuck to hitting the rocks with
AK> a hammer. Or a ton of explosives and a drilling bit <G>.

fair point, but I'm hammering only my keyboard, no rocks.

AK> I guess I'll have to go and look at the KTB website now, to see
AK> if there's been anything since my last visit.

AK> --
AK> Aidan Karley,
AK> Aberdeen, Scotland,
AK> Location: 57°10'11" N, 02°08'43" W (sub-tropical Aberdeen), 0.021233

--
ingo wardinski GeoForschungsZentrum Potsdam, in...@gfz-potsdam.de

**********************************************************************
"It's your money. You paid for it." George W. Bush
LaCrosse, Wis., Oct. 18, 2000

[Hans-Bernhard Broeker]

Ingo Wardinski <in...@gfz-potsdam.de> wrote:
> Hi,
> I would like to render the attached data in a logarithmic style, but
> as there are some negative values, the plot looks sketchy.

Let's forget "looks", and discuss rationales instead. Given the fact
they're completely unsuited for being interpreted that way, what on
earth makes you want to plot these data on a log scale? Note that
with the suggested "fix"

plot 'data' using 1:($2>0?log($2):-log(-$2))

your plot would display both y = -1000.0 and y = +0.001 as identical
"-3" --- I strongly doubt that's what you wanted.

The only sane way of plotting these data with a log y axis I can see
would be to split the dataset into one for the negative y values and
one for the positive, e.g. like this:

plot 'data' using 1:($2>0:$2:0/0) title 'y>0', \
'' using 1:($2<0?-$2:0/0) title 'y<0'

Or you could use multiplot mode to put them into separate diagrams,
one with an upward log y axis for the positive y values, the other
with a downward log axis used for the negative values only.

--
Hans-Bernhard Broeker (bro...@physik.rwth-aachen.de)
Even if all the snow were burnt, ashes would remain.