Fwd: Accurate probability distribution functions (GSOC 2012)
michael.baudin at contrib.scilab.org
michael.baudin at contrib.scilab.org
Mon Mar 26 15:54:10 CEST 2012
The Inverse Beta is implemented here :
http://forge.scilab.org/index.php/p/distfun/source/tree/HEAD/macros/distfun_betainv.sci
This is currently just a macro which calls cdfbet with consistent
arguments. One possible improvement is to refactor the function into a
proper C gateway, with much more flexibility that the current cdfbet
function. For example, betainv can take as arguments a 1-by-1 double "a"
parameter and a 5-by-6 matrix "b" parameter: in this case, the parameter
"a" is expanded into a 5-by-6 matrix. The current cdfbet function cannot
do this.
It is a good thing that you are interested in the Geometric
distribution, since it is not currently implemented. As an exercise,
please create a geopdf function on the model of the Matlab function :
http://www.mathworks.fr/help/toolbox/stats/geopdf.html
Once done, please send your script to us. Please do not spend more that
a few hours on it. A script with comments is unnecessary at this point:
just focus on the Scilab code. More that 100 lines would be very
surprising.
When you are done, please pause on the research phase and fill-in your
application form. We are currently in the early stages of the selection
process: you will have plenty of time if you get selected.
Best regards,
Michaël Baudin
On Mon, 26 Mar 2012 03:36:09 +0530, prateek papriwal
<papriwalprateek at gmail.com> wrote:
> Where is the inverse CDF's implemented (Beta inverse cdf,normal
> inverse cdf etc. as menationed in [1]) . The list of functions(given
> at link [2]) under sub heading Statistics does not have these
> functions . Also i looked upon the functions there can be a lot more
> can be implemented . Matlab has huge number of functions .
>
> One of such functions as Geometric CDF , i have planned for how to go
> about implenting it. Geometric CDF is the probability distribution of
> the number of bernoulli trials to get one success.
>
> Calling Sequence
>
> [P,Q] = cdfgeo("PQ",Xn,Pr,Ompr)
>
> Arguments
> P,Q,Xn,PR,Ompr
> 5 real vecors of the same size
>
> P,Q(Q=1-P)
> The cumulation from 0 to Xn of the geometric distribution.
> (Probablility of getting one(first) success in Xn trials or
> less than Xn. Probability of success is Pr . ) Input range: [0,1].
>
> Xn
> The number of max trials .We will be looking at the geometric
> distribution for 1 to Xn trials and then sum them up to get geocdf.
>
> Pr,Ompr
> The probability of getting success in each trial .
>
> In matlab it is implemented as
>
> geocdf(3,0.5) where 3 is our Xn and Pr=Ompr=0.5
>
> Hence now geocdf will be = (0.5^0)(0.5) + ((0.5)^1)(0.5) +
> ((0.5)^2)(0.5) = 0.5 + 0.25 + 0.125
> Explanation of 3rd term -- to get first success, i used 3 trials(Xn)
> , therefore got failure in first two trials and success in the 3rd
> trial .
> Explanation of 2nd Term -- to get first success , i used 2 trials
> (less than Xn) , therefore got failure in first trial and success in
> 2nd one...
> Explanation of 1st Term -- i got success in first trial only ...
>
> Note -- Xn is the total number of max trials. For determining cdf we
> will be using fewer number tan Xn also and adding all them up ..
>
> [1] http://forge.scilab.org/index.php/p/distfun/ [1]
> [2] http://help.scilab.org/docs/5.4.0-alpha-1/en_US/index.html [2]
>
> On Sat, Mar 24, 2012 at 9:50 PM, Sylvestre Ledru wrote:
> thanks for the forward, just don't forget me the next time ;)
>
> Le samedi 24 mars 2012 à 17:19 +0530, prateek papriwal a écrit :
>
>>
> >
> > ---------- Forwarded message ----------
> > From: prateek papriwal
> > Date: Sat, Mar 24, 2012 at 4:47 PM
> > Subject: Re: Accurate probability distribution functions (GSOC
> 2012)
> > To: michael.baudin at contrib.scilab.org [5]
> >
> >
> > ok .
> >
> > yes . its true there are much more functions in matlab.
> > PLUS more function such as
> > binmial,geometric,hypergeometric,poisson,chebyshev's inequality,
> > markov's inequality, MGF of random variable , chi- square,
> bernoulli
> > etc.
> >
> > Yes implementing these functions in scilab would be a good idea to
> > work upon .
> >
> >
> > On Sat, Mar 24, 2012 at 4:27 PM,
> > wrote:
> > I think that you looked at it too quickly.
> > There is, by far, not all the required functions.
> > To see this more quickly, compare with :
> >
> >
> http://www.mathworks.fr/help/toolbox/stats/bq_w_hm.html#bq_w_ie-8 [7]
> >
> > There is only ~5% of the distributions in Matlab.
> > At least 30-40 more functions are necessary to make a
> complete
> > distribution toolbox:
> > just look at the "TODO" section in the readme.
> >
> > As far as I am concerned, I have no interest in Queuing
> > theory.
> >
> > Regards,
> >
> > Michaël
> >
> >
> >
> > On Mon, 19 Mar 2012 18:05:25 +0530, prateek papriwal
> > wrote:
> >
> > @michael, thank you for ur reply .
> >
> >
> >
> http://forge.scilab.org/index.php/p/distfun/ [9] [1] .. i
> > saw the
> >
> > features(functions) already implemented .
> > i would like to have a view at the current
> state and
> > would be very
> > mjuch interested in implemeting concepts of
> Queueing
> > theory .
> >
> >
> > On Sun, Mar 18, 2012 at 7:24 AM, Sylvestre
> Ledru
> > wrote:
> >
> > Le samedi 17 mars 2012 à 17:24 +0100,
> >
> > michael.baudin at contrib.scilab.org [10] [3]
> >
> > a écrit :
> >
> >
> > > Secondly, i now work at EDF R&D, and
> no more in
> > the Scilab
> > > Consortium.
> > > How i could be involved in the GSOC
> 2012 has to be
> > clarified with
> > > the
> > > Consortium.
> >
> > As long as I (or someone else in the
> consortium) am
> > in the loop, I
> > don't
> > have any issues with that.
> >
> > S
> >
> > --
> > -----------------------------
> > Sylvestre Ledru
> > Projects manager
> > Community manager
> > -----------------------------
> > Scilab Enterprises
> >
> > http://www.scilab-enterprises.com/ [11]
> [4]
> > http://www.scilab.org/ [12] [5]
> > -----------------------------
> >
> >
> >
> > Links:
> > ------
> > [1]
> http://forge.scilab.org/index.php/p/distfun/ [13]
> > [2]
> mailto:sylvestre.ledru at scilab-enterprises.com [14]
> > [3]
> mailto:michael.baudin at contrib.scilab.org [15]
> > [4] http://www.scilab-enterprises.com/ [16]
> > [5] http://www.scilab.org/ [17]
> >
> >
> >
>
> --
> -----------------------------
> Sylvestre Ledru
> Operation manager
>
> Community manager
> -----------------------------
> Scilab Enterprises
> http://www.scilab-enterprises.com/ [18]
> http://www.scilab.org/ [19]
> -----------------------------
>
>
>
> Links:
> ------
> [1] http://forge.scilab.org/index.php/p/distfun/
> [2] http://help.scilab.org/docs/5.4.0-alpha-1/en_US/index.html
> [3] mailto:sylvestre.ledru at scilab-enterprises.com
> [4] mailto:papriwalprateek at gmail.com
> [5] mailto:michael.baudin at contrib.scilab.org
> [6] mailto:michael.baudin at contrib.scilab.org
> [7] http://www.mathworks.fr/help/toolbox/stats/bq_w_hm.html#bq_w_ie-8
> [8] mailto:papriwalprateek at gmail.com
> [9] http://forge.scilab.org/index.php/p/distfun/
> [10] mailto:michael.baudin at contrib.scilab.org
> [11] http://www.scilab-enterprises.com/
> [12] http://www.scilab.org/
> [13] http://forge.scilab.org/index.php/p/distfun/
> [14] mailto:sylvestre.ledru at scilab-enterprises.com
> [15] mailto:michael.baudin at contrib.scilab.org
> [16] http://www.scilab-enterprises.com/
> [17] http://www.scilab.org/
> [18] http://www.scilab-enterprises.com/
> [19] http://www.scilab.org/
More information about the users
mailing list