You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@commons.apache.org by fr...@bredband.net on 2004/10/12 20:58:34 UTC

[math] Questions regarding probability distributions

Hi

I've been looking around for open source mathematical statistics software in 
Java in the last couple of weeks and have tested the commons-math package.
After looking into it I have a few questions and comments.

1. The MathUtils.factorial(final int n) method throws an 
IllegalArgumentException for n=0 which is clearly incorrect since 0!=1 by 
definition.

2. I think the distinction between discrete and continuous probability 
distributions manifested in the interfaces DiscreteDistribution and 
ContinuousDistribution is rather questionable There is no requirement that 
a discrete distribution only takes on integer values so the methods of 
interface DiscreteDistribution doesn't cover all discrete distributions. On  
the other hand, all of the methods of ContinuousDistribution holds equally 
well for a discrete probability distribution. In my opinion, a more 
appropriate approach would be to rename ContinuousDistribution to 
ProbabilityDistribution and drop the DiscreteDistribution interface. Of 
course, it could be practical to have  convenience methods that takes integer 
arguments for the probability densities for certain distributions but then 
you can define a new interface IntegerValuedDistribution like 

public interface IntegerValuedDistribution extends ProbabilityDistribution  {
    double probability(int x);
    double cumulativeProbability(int x) throws MathException;
}

3. Wouldn't it be nice to have convenience methods for calculating
the moments for each distribution? Something along the lines of

public double getMoment(int order) throws MomentDoesNotExistException;

for calculating the moments EX^order (if they exist)? 
At least there could be methods for obtaining the mean and variance of a 
distribution.

4.  Since the chi-squared and exponential distributions are just special cases
of the gamma distribution, there is no need to have separate implementation
classes for these. In my opinion, one should avoid having multiple 
implementations of the same distribution unless there is some strong reason 
for it.

5. There are quite a lot of elementary distributions missing.
I wrote an implementation of the Poisson distribution while testing the 
package and have attached the files for it.

Best regards,
Frank N

Re: [math] Questions regarding probability distributions

Posted by "Mark R. Diggory" <md...@latte.harvard.edu>.

frno@bredband.net wrote:
>>>4.  Since the chi-squared and exponential distributions are just special
>>>cases of the gamma distribution, there is no need to have separate
>>>implementation classes for these. In my opinion, one should avoid having
>>>multiple implementations of the same distribution unless there is some
>>>strong reason for it.
>>
>>Please keep in mind that not every user is a full-blown statistician
>>knowing all these underlying interactions. For that reason, I would
>>argue strongly against this.
>>
> 
> 
> It wouldn't make any difference to the end user as I guess he's supposed to 
> use the DistributionFactory to create the distributions he need.
> 
> You could keep the ExponentialDistribution and ChiSquaredDistribution 
> interfaces and just change the following in DistributionFactoryImpl:
> 
> public ChiSquaredDistribution createChiSquareDistribution(
>         final double degreesOfFreedom) {
>  
>  return new GammaDistributionImpl(degreesOfFreedom/2, 2);
> }
> 
> public ExponentialDistribution createExponentialDistribution(double mean) {
>         
>  return new GammaDistributionImpl(1, mean);
> }

Very Bad Idea, this means that GammaDistribution would have to implement 
Exponential and ChiSquare interfaces, if the "beta scale parameter" 
exposed in the Gamma interface were changed then this would be 
exceptionally confusing to the user. They could in a sense make a 
ChiSquare instance into some other distribution by changing the 
properties available on the class. This is what the default 
implementation of ChiSquareDist should be enforcing, if it is not, then 
the interface should be further restricted.

> 
> Then you can drop ExponentialDistributionImpl and ChiSquaredDistributionImpl 
> altogether.

The way to do this is to maintain ExponentialDistributionImpl and 
ChiSquaredDistributionImpl as "wrappers" that just lockdown 
GammaDistribution to a specific configuration (via special Constructor 
and overriding the api to not allow the "beta scale parameter" to change).

This is the point of these interfaces/implementations. Yes, everyone 
could learn that these are all just variations on a Gamma distribution, 
I learned that in my college statistics class. But providing common 
default implementations is important for ease of use.

-Mark

-- 
Mark Diggory
Software Developer
Harvard MIT Data Center
http://www.hmdc.harvard.edu

---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by "Mark R. Diggory" <md...@latte.harvard.edu>.
I definitely against removing these implementations. These 
interfaces/implementations provide a a unique set of methods on that 
specific species of Gamma distribution that are required to configure 
and maintain that species of Gamma distribution.

Removing the underlying implementations and making a 
GammaDistributionImpl implement the ChiSquareDistibution interface is 
risky and difficult venture, there would be no way to enforce the 
"immutability" of the scale parameter required to maintain its 
"ChiSquareness".

-Mark

Kim van der Linde wrote:
> 
> 
> Phil Steitz wrote:
> 
>> Good point. I am +0 for dropping ChiSquaredDistributionImpl, but -1 
>> for exponential (because of efficiency).  Unless others object, I will 
>> make that change.  Thanks for pointing this out.
> 
> 
> I do not mind HOW it is in the package, but I would argue against an 
> option where the user has to know which parameters need to be fed to an 
> more general underlying distribution, especially when it is a common 
> used one such as chi-square. Unless of course, you want to limit this 
> package to a hard core public only.
> 
> Cheers,
> 
> Kim

-- 
Mark Diggory
Software Developer
Harvard MIT Data Center
http://www.hmdc.harvard.edu

---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by Kim van der Linde <ki...@kimvdlinde.com>.

Phil Steitz wrote:

> Good point. I am +0 for dropping ChiSquaredDistributionImpl, but -1 for 
> exponential (because of efficiency).  Unless others object, I will make 
> that change.  Thanks for pointing this out.

I do not mind HOW it is in the package, but I would argue against an 
option where the user has to know which parameters need to be fed to an 
more general underlying distribution, especially when it is a common 
used one such as chi-square. Unless of course, you want to limit this 
package to a hard core public only.

Cheers,

Kim
-- 
http://www.kimvdlinde.com

---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by Phil Steitz <ph...@steitz.com>.
frno@bredband.net wrote:
> 
> It wouldn't make any difference to the end user as I guess he's supposed to 
> use the DistributionFactory to create the distributions he need.
> 
> You could keep the ExponentialDistribution and ChiSquaredDistribution 
> interfaces and just change the following in DistributionFactoryImpl:
> 
> public ChiSquaredDistribution createChiSquareDistribution(
>         final double degreesOfFreedom) {
>  
>  return new GammaDistributionImpl(degreesOfFreedom/2, 2);
> }
> 
> public ExponentialDistribution createExponentialDistribution(double mean) {
>         
>  return new GammaDistributionImpl(1, mean);
> }
> 
> Then you can drop ExponentialDistributionImpl and ChiSquaredDistributionImpl 
> altogether.
> 
> I guess this is no big deal now but in general one should avoid having 
> multiple implementations for the same distribution since that increases the 
> risk for errors. In this case you could possibly argue that there is a small 
> performance gain in having separate implementations, at least for the 
> ExponentialDistribution.

Good point. I am +0 for dropping ChiSquaredDistributionImpl, but -1 for 
exponential (because of efficiency).  Unless others object, I will make 
that change.  Thanks for pointing this out.

Phil
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: commons-dev-help@jakarta.apache.org
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by fr...@bredband.net.
> > 4.  Since the chi-squared and exponential distributions are just special
> > cases of the gamma distribution, there is no need to have separate
> > implementation classes for these. In my opinion, one should avoid having
> > multiple implementations of the same distribution unless there is some
> > strong reason for it.
>
> Please keep in mind that not every user is a full-blown statistician
> knowing all these underlying interactions. For that reason, I would
> argue strongly against this.
>

It wouldn't make any difference to the end user as I guess he's supposed to 
use the DistributionFactory to create the distributions he need.

You could keep the ExponentialDistribution and ChiSquaredDistribution 
interfaces and just change the following in DistributionFactoryImpl:

public ChiSquaredDistribution createChiSquareDistribution(
        final double degreesOfFreedom) {
 
 return new GammaDistributionImpl(degreesOfFreedom/2, 2);
}

public ExponentialDistribution createExponentialDistribution(double mean) {
        
 return new GammaDistributionImpl(1, mean);
}

Then you can drop ExponentialDistributionImpl and ChiSquaredDistributionImpl 
altogether.

I guess this is no big deal now but in general one should avoid having 
multiple implementations for the same distribution since that increases the 
risk for errors. In this case you could possibly argue that there is a small 
performance gain in having separate implementations, at least for the 
ExponentialDistribution.


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by Kim van der Linde <ki...@kimvdlinde.com>.

frno@bredband.net wrote:
> Hi

> 4.  Since the chi-squared and exponential distributions are just special cases
> of the gamma distribution, there is no need to have separate implementation
> classes for these. In my opinion, one should avoid having multiple 
> implementations of the same distribution unless there is some strong reason 
> for it.

Please keep in mind that not every user is a full-blown statistician 
knowing all these underlying interactions. For that reason, I would 
argue strongly against this.

Kim

-- 
http://www.kimvdlinde.com

---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by F Norin <fr...@bredband.net>.
> > Well, the problem is this: If I need to create some custom discrete
> > distribution that doesn't take on integer values, what interface should I
> > implement? With your model I have no choice but use the
> > ContinuousDistribution interface even though the distribution *isn't*
> > continuous. Does that make sense?
>
> Can you provide a practical example of this?  IIUC, what you are really
> arguing for is changing the int's in the DiscreteDistribution interface to
> doubles. This has the advantage of greater generality but makes it
> slightly less convenient for implementors of the most common discrete
> distributions, where the values are integers.

Well, changing the int's in the DiscreteDistribution interface to doubles is 
kind of a workaround, but I don't think it will settle the issue for good,  
see below.

As for examples, you can take *any* mixed distribution as an example of what I
mean. Consider a random variable X with domain D that can be partitioned
into subsets A and B such that

1. A is a countable set and 0 < P(X is in A) < 1
2. P(X = x) = 0 for all x in B

How would the distribution for such a random variable be represented in
your framework?

As a simple example of this, consider a random variable with the density

f(x) = 0.5 for x=0
f(x) = 0.5 for 1<x<2

How does this distribution fit into your framework? Sure, you could have
it implement the ContinuousDistribution interface but it *isn't* a
continuous distribution (in the sense that it doesn't conform to the
definition of a continuous distribution in probability theory) - and
then it shouldn't implement an interface called ContinuousDistribution.

Recall: A random variable is continuous if its distribution function P(X <= x) 
can be expressed as the Riemann-integral of some integrable function 
f: R -> [0, infinity)

The basic problem is that you have an implicit assumption in your
framework that each and every probability distribution can be classified
as being either discrete or continuous . That is simply not true.
Discrete and continuous distributions are really only special cases of
a broader concept. Aside from that you also have the problem of how to
handle the case of a discrete distribution that doesn't take on integer
values.

Note: There are also distributions that are neither discrete, continuous or a 
mixture of the two. For example, there are numerous distributions based upon  
the Cantor ternary sets.

The bottom line is that you *cannot* do without a generic
ProbabilityDistribution interface.
This interface should expose a method that exists for all and completely
determines a particular probability distribution, such as the
distribution function P(X <= x).

As an easy solution, you could define it as

public interface ProbabilityDistribution {
        public double distributionFunction(double x);
}

and have ContinuousDistribution and DiscreteDistribution extend it.

This should work ok (though the name DiscreteDistribution is misleading)
but if you want a completely generic and typesafe definition you should
go for something like

public interface ProbabilityDistribution {
        public Probability distributionFunction(Number x);
}

where Number is the standard java.lang.Number and Probability is a new
class that would need to be defined. Since probability measure is the
fundamental concept in probability theory, having a Probability class is
probably a good idea anyway. It could look something like this

public final class Probability implements Serializable {

    public static final Probability PROPABILITY_ONE = new
Probability(1);
    public static final Probability PROPABILITY_ZERO = new
Probability(0);

    private double value;

    public Probability(double v) throws IllegalArgumentException {
        if (v < 0 || v > 1) {
            throw new IllegalArgumentException("Illegal probability
value.");
        }
        this.value = v;
    }

    public double value() {
        return value;
    }
}

(A custom equals method should also be provided of course.)


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by Phil Steitz <ph...@steitz.com>.
frno@bredband.net wrote:
>>Some may disagree with that definition (really amounts to a convention on
>>how to handle vacuuous products); but I think that I agree with you.
>>Unless others object, I will consider this a bug and make the change.
>>Thanks for pointing this oddity out. Would you mind opening a bugzilla
>>ticket to track this?
> 
> 
> Fixed that.

Thanks!
> 

> 
> Well, the problem is this: If I need to create some custom discrete 
> distribution that doesn't take on integer values, what interface should I 
> implement? With your model I have no choice but use the 
> ContinuousDistribution interface even though the distribution *isn't* 
> continuous. Does that make sense?

Can you provide a practical example of this?  IIUC, what you are really 
arguing for is changing the int's in the DiscreteDistribution interface to 
doubles. This has the advantage of greater generality but makes it 
slightly less convenient for implementors of the most common discrete 
distributions, where the values are integers.

> 
> You're right that the domain of every discrete distribution could be mapped 
> into the integers, but then there should be some mechanism for internalizing 
> this mapping for a particular distribution. Every discrete distribution 
> should have a reference to a Domain object that provides a method for mapping 
> every object of the set into the integers. Using such a scheme you could keep 
> separate interfaces for discrete and continuous distributions. (Actually, the 
> continuous distributions could have a Domain object too).

Hmm, one problem that I am now seeing is in computing moments of "mapped" 
discrete distributions. Something like what you are describing above will 
be necessary.  So...if there are practical examples out there, I am 
beginning to think that we need to either a) change the ints to doubles in 
DiscreteDistribution or b) support "value mapping" in some systematic way.

Phil
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by fr...@bredband.net.
> Some may disagree with that definition (really amounts to a convention on
> how to handle vacuuous products); but I think that I agree with you.
> Unless others object, I will consider this a bug and make the change.
> Thanks for pointing this oddity out. Would you mind opening a bugzilla
> ticket to track this?

Fixed that.


> > There is no requirement that
> > a discrete distribution only takes on integer values so the methods of
> > interface DiscreteDistribution doesn't cover all discrete distributions.
>
> Technically, you are correct - a discrete random variable can take on any
> (countable) set of values. The key difference, however, is that for a
> discrete random variable X, if x is one of the values that it takes with
> positive probability, p(X=x) is non-zero. Therefore, it makes sense to
> have probability(x) defined (as it is) in the DiscreteDistribution
> interface, but absent from the ContinuousDistribution interface.  More
> care also needs to be taken in the discrete case to define and implement
> p(a < X < b) type quantities including or excluding endpoints
> appropriately (since it makes a difference whether or not endpoints are
> included).
>
> In fairness, the DiscreteDistribution interface only supports
> integer-valued distributions. The only ones that I have ever used are
> integer-valued, and any odd countable set can be mapped into the integers,
> so I don't personally view this as a serious limitation.
>
>   On
>
> > the other hand, all of the methods of ContinuousDistribution holds
> > equally well for a discrete probability distribution.
>
> With slightly different semantics, as noted above.
>
>   In my opinion, a more
>
> > appropriate approach would be to rename ContinuousDistribution to
> > ProbabilityDistribution and drop the DiscreteDistribution interface. Of
> > course, it could be practical to have  convenience methods that takes
> > integer arguments for the probability densities for certain distributions
> > but then you can define a new interface IntegerValuedDistribution like
> >
> > public interface IntegerValuedDistribution extends
> > ProbabilityDistribution  { double probability(int x);
> >     double cumulativeProbability(int x) throws MathException;
> > }
>
> I see your point here and it is legitimate.  I am -0 to making the change,
> however, since the interface contract (for the base interface) would in
> practical terms be different for the discrete case, so I think it is
> better to keep them separate.

Well, the problem is this: If I need to create some custom discrete 
distribution that doesn't take on integer values, what interface should I 
implement? With your model I have no choice but use the 
ContinuousDistribution interface even though the distribution *isn't* 
continuous. Does that make sense?

You're right that the domain of every discrete distribution could be mapped 
into the integers, but then there should be some mechanism for internalizing 
this mapping for a particular distribution. Every discrete distribution 
should have a reference to a Domain object that provides a method for mapping 
every object of the set into the integers. Using such a scheme you could keep 
separate interfaces for discrete and continuous distributions. (Actually, the 
continuous distributions could have a Domain object too).


> > 4.  Since the chi-squared and exponential distributions are just special
> > cases of the gamma distribution, there is no need to have separate
> > implementation classes for these. In my opinion, one should avoid having
> > multiple implementations of the same distribution unless there is some
> > strong reason for it.
>
> Well, the chi-square implementation wraps a gamma instance.  The
> exponential implementation computes the exponential density explicitly.
> In any case, the distributions are different (though related) and
> sufficiently useful in their own rights that exposing them separately will
> be easier for users.

See my answer to another post.


> > 5. There are quite a lot of elementary distributions missing.
> > I wrote an implementation of the Poisson distribution while testing the
> > package and have attached the files for it.
>
> Yes!! That is why the framework was designed to support adding
> distributions. Could you open a bugzilla ticket and attach the files
> there?  I will review and get these in ASAP.

Fixed that.


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by Al Chou <ho...@yahoo.com>.
--- Phil Steitz <ph...@steitz.com> wrote:
> Thanks for the feedback and the contribution!
> 
> See comments interspersed below.
> 
> frno@bredband.net wrote:
> > Hi
> > 
> > I've been looking around for open source mathematical statistics software
> in 
> > Java in the last couple of weeks and have tested the commons-math package.
> > After looking into it I have a few questions and comments.
> > 
> > 1. The MathUtils.factorial(final int n) method throws an 
> > IllegalArgumentException for n=0 which is clearly incorrect since 0!=1 by 
> > definition.
> 
> Some may disagree with that definition (really amounts to a convention on 
> how to handle vacuuous products); but I think that I agree with you. 
> Unless others object, I will consider this a bug and make the change. 
> Thanks for pointing this oddity out. Would you mind opening a bugzilla 
> ticket to track this? 
>
<http://issues.apache.org/bugzilla/enter_bug.cgi?product=Commons&component=Math>

factorial( 0 ) === 1 by definition, as stated above.  Any definition of
factorial that differs from that is not mathematically correct and will
surprise those who expect the mathematical definition.  So, +1 to making the
change.


> > 3. Wouldn't it be nice to have convenience methods for calculating
> > the moments for each distribution? Something along the lines of
> > public double getMoment(int order) throws MomentDoesNotExistException;
> > 
> > for calculating the moments EX^order (if they exist)? 
> > At least there could be methods for obtaining the mean and variance of a 
> > distribution.
> 
> Yes!!  We thought about adding these to the interfaces but were concerned 
> that in some cases they would be hard to implement (or impossible -- if 
> they don't exist ;-)  There is nothing preventing us from adding them to 
> the implementations, however, or adding another interface or utilities to 
> compute them. Does anyone see a reason that we need to add them now, 
> before releasing 1.0?

Not being a statistician, I think I can only vote -0 (not -1) to adding them at
this time, in good conscience.


> > 4.  Since the chi-squared and exponential distributions are just special
> cases
> > of the gamma distribution, there is no need to have separate implementation
> > classes for these. In my opinion, one should avoid having multiple 
> > implementations of the same distribution unless there is some strong reason
> 
> > for it.
> 
> Well, the chi-square implementation wraps a gamma instance.  The 
> exponential implementation computes the exponential density explicitly. 
> In any case, the distributions are different (though related) and 
> sufficiently useful in their own rights that exposing them separately will 
> be easier for users.

I agree they should remain separate.  Although we have yet to really see who
our users are, it seems reasonable, given Commons Math's charter, to keep ease
of comprehension/use as one of our goals.  Mathematical sophistication should
not be a prerequisite (or lack thereof present a barrier) to entry into using
this library.



Al

---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Re: [math] Questions regarding probability distributions

Posted by Phil Steitz <ph...@steitz.com>.
Thanks for the feedback and the contribution!

See comments interspersed below.

frno@bredband.net wrote:
> Hi
> 
> I've been looking around for open source mathematical statistics software in 
> Java in the last couple of weeks and have tested the commons-math package.
> After looking into it I have a few questions and comments.
> 
> 1. The MathUtils.factorial(final int n) method throws an 
> IllegalArgumentException for n=0 which is clearly incorrect since 0!=1 by 
> definition.

Some may disagree with that definition (really amounts to a convention on 
how to handle vacuuous products); but I think that I agree with you. 
Unless others object, I will consider this a bug and make the change. 
Thanks for pointing this oddity out. Would you mind opening a bugzilla 
ticket to track this? 
<http://issues.apache.org/bugzilla/enter_bug.cgi?product=Commons&component=Math>

> 
> 2. I think the distinction between discrete and continuous probability 
> distributions manifested in the interfaces DiscreteDistribution and 
> ContinuousDistribution is rather questionable.

Here I don't think I agree.  Continuous and discrete distributions are 
fundamentally different. See below.

> There is no requirement that 
> a discrete distribution only takes on integer values so the methods of 
> interface DiscreteDistribution doesn't cover all discrete distributions.

Technically, you are correct - a discrete random variable can take on any 
(countable) set of values. The key difference, however, is that for a 
discrete random variable X, if x is one of the values that it takes with 
positive probability, p(X=x) is non-zero. Therefore, it makes sense to 
have probability(x) defined (as it is) in the DiscreteDistribution 
interface, but absent from the ContinuousDistribution interface.  More 
care also needs to be taken in the discrete case to define and implement 
p(a < X < b) type quantities including or excluding endpoints 
appropriately (since it makes a difference whether or not endpoints are 
included).

In fairness, the DiscreteDistribution interface only supports 
integer-valued distributions. The only ones that I have ever used are 
integer-valued, and any odd countable set can be mapped into the integers, 
so I don't personally view this as a serious limitation.

  On
> the other hand, all of the methods of ContinuousDistribution holds equally 
> well for a discrete probability distribution.

With slightly different semantics, as noted above.

  In my opinion, a more
> appropriate approach would be to rename ContinuousDistribution to 
> ProbabilityDistribution and drop the DiscreteDistribution interface. Of 
> course, it could be practical to have  convenience methods that takes integer 
> arguments for the probability densities for certain distributions but then 
> you can define a new interface IntegerValuedDistribution like 
> 
> public interface IntegerValuedDistribution extends ProbabilityDistribution  {
>     double probability(int x);
>     double cumulativeProbability(int x) throws MathException;
> }

I see your point here and it is legitimate.  I am -0 to making the change, 
however, since the interface contract (for the base interface) would in 
practical terms be different for the discrete case, so I think it is 
better to keep them separate.

> 
> 3. Wouldn't it be nice to have convenience methods for calculating
> the moments for each distribution? Something along the lines of
> public double getMoment(int order) throws MomentDoesNotExistException;
> 
> for calculating the moments EX^order (if they exist)? 
> At least there could be methods for obtaining the mean and variance of a 
> distribution.

Yes!!  We thought about adding these to the interfaces but were concerned 
that in some cases they would be hard to implement (or impossible -- if 
they don't exist ;-)  There is nothing preventing us from adding them to 
the implementations, however, or adding another interface or utilities to 
compute them. Does anyone see a reason that we need to add them now, 
before releasing 1.0?

> 4.  Since the chi-squared and exponential distributions are just special cases
> of the gamma distribution, there is no need to have separate implementation
> classes for these. In my opinion, one should avoid having multiple 
> implementations of the same distribution unless there is some strong reason 
> for it.

Well, the chi-square implementation wraps a gamma instance.  The 
exponential implementation computes the exponential density explicitly. 
In any case, the distributions are different (though related) and 
sufficiently useful in their own rights that exposing them separately will 
be easier for users.

> 
> 5. There are quite a lot of elementary distributions missing.
> I wrote an implementation of the Poisson distribution while testing the 
> package and have attached the files for it.

Yes!! That is why the framework was designed to support adding 
distributions. Could you open a bugzilla ticket and attach the files 
there?  I will review and get these in ASAP.

Thanks again for the feedback and contribution.

Phil

---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org