Re: [Gnuplot-info] Log scale not working?

[zweifel <zw...@gm...>]

How do I supply data errors for fit to work with?

2010/5/14 Hans-Bernhard Bröker <HBB...@t-...>

> zweifel wrote:
>
>  I plotted a function with a regression and it worked fine as it could be
>> seen on the figure out2.png, but when I tried to plot this same function
>> with logscale, that is, I just added "set logscale xy" to my script, then
>> the out.png appeared. But it did not seem right.
>>
>
> On what basis did you draw this conclusion?  What exactly did "not seem
> right" about it?
>
>
>  fit f(x) "comparison" using 1:2 via e,a
>> fit g(x) "comparison" using 1:3 via d,k2
>> fit h(x) "comparison" using 1:4 via c,k3
>>
>
> The most likely problem to be found here is that these fit commands don't
> supply data errors for fit to work with.  Your expectation is probably based
> on assumptions about how 'fit' weights distances between data and model
> (particularly in the absence of data errors) that just don't hold.
>
> Note that 'set logscale' is a plot setting.  It does not affect 'fit at
> all.
>
>


-- 

Danilo