Re: [Gnuplot-info] plot distribution

[Thomas S. <t.s...@fz...>]

the normalization factor needs to be know before you plot.

what you could do:
- make a dummy plot (see 'set table') to count the data entries,
  have a look at:
  http://gnuplot.sourceforge.net/demo_4.3/data_feedback.html
  (look for 'sum')
- calculate the normalization factor
- do the plot

e.g.:
set table
sum=0
plot 'datafile' using 1:(sum=sum+$1)
unset plot
print sum
plot 'datafile' using 1:(1./sum) smooth frequency with boxes


jamborta wrote:
> 
> that's great. it's clear now. maybe it'd be useful to have a function that
> counts how many entries the set has, so it'd be easy to normalise it.
> 
> 
> Thomas Sefzick wrote:
>> 
>> first we need a function for binning the x-data, i.e. the
>> continuous x-range is split into intervals of size 's':
>> bin(x, s) = s*int(x/s)
>> 
>> then we plot:
>> plot "datafile" using (bin($2,0.01)):(10/300.) \
>>    smooth frequency \
>>    title 'smooth frequency' with boxes
>> 
>> explanation:
>> - the bin-function assigns the x-values within one interval
>>   (which has a width of 0.01 here) to only one x-value (the
>>   lower interval border)
>> - with 'smooth frequency' all the y-values of data pairs
>>   with the same x-value are added
>> - "10/300." is just a factor, you could use "1." then each
>>   data counts as '1'
>> 
>> 
>> jamborta wrote:
>>> 
>>> thanks a lot. could you explain the parameters of this method
>>> 
>>> plot [0:] "datafile" u 2:(0.25*rand(0)-.35) t '', "" u
>>> (bin($2,0.01)):(10/300.) s f t 'smooth frequency' w boxes, "" u
>>> 2:(1/300.) s cumul t 'smooth cumulative'
>>> 
>>> especially 10/300. and 1/300.
>>> 
>>> I couldn't figure out why it normalises in the demo, but not for my
>>> data.
>>> 
>>> thanks a lot
>>> 
>>> 
>>> Thomas Sefzick wrote:
>>>> 
>>>> have a look at the gnuplot 4.3-cvs demo:
>>>> 
>>>> http://gnuplot.sourceforge.net/demo_4.3/smooth.html
>>>> 
>>>> 
>>>> jamborta wrote:
>>>>> 
>>>>> hi guys,
>>>>> 
>>>>> I have a continuous random variable with around 1200 values. I'd like
>>>>> to plot the distribution of that dataset. I guess I could set up some
>>>>> intervals and count how many values fall into that interval. is that
>>>>> doable with gnuplot?
>>>>> 
>>>>> thanks,
>>>>> tamas
>>>>> 
>>>>> 
>>>> 
>>>> 
>>> 
>>> 
>> 
>> 
> 
> 

-- 
View this message in context: http://www.nabble.com/plot-distribution-tp24593401p24620723.html
Sent from the Gnuplot - User mailing list archive at Nabble.com.