Hi everyone,
David King wrote:
> May I add these suggestions to the list?
>
> B function
> * "this function returns the probability that the number of successful
> trials shall lie between S and S2." - should say clearly if inclusive or
> exclusive (inclusive I think).
> * Test cases: " =B(7;0.5;3) 0.24343750" - wrong result - 0.2734375
> is correct, as returned by Calc, and checked from the actual formula by
> me.
>
confirmed in R 2.6.2
> dbinom(3, 7, 0.5)
[1] 0.2734375
> pbinom(7, 8, 0.2)-pbinom(2, 8, 0.2)
[1] 0.2030797
(NOT 0.20307698)
On a different note, I noticed that OOo Calc and other spreadsheet
applications really mix up the density and the distribution functions:
e.g. B() does not have specifically a (cumulative) distribution
function. Of course B(x, p, 0 , k) will compute the distribution function.
Wouldn't it be advisable to define *ALL density and ALL distribution
functions* like:
SOME_FUNCTION(..., distribution = FALSE)
SOME_FUNCTION(..., distribution = TRUE)
for the 2 corresponding functions.
Also, use consistently then the 'distribution'-keyword inside the
Open-Formula document to specify cumulative probabilities (distribution).
Then, B(x, p, k1, k2) could be computed using the new function as:
B(x, p, k2, dist = TRUE) - B(x, p, k1 - 1, dist = TRUE)
['k1 - 1' IF both k1 & k2 are included - as requested by David, this
should be specified]
While some may complain about the greater flexibility of the existing
implementation, the truth is, i.) this is equally flexible and ii.)
this is more consistent.
I especially like R's handling of the issue: ALL distribution types have
the following functions defined (as separate functions):
Density (d), distribution function (p), quantile function (q) and random
generation (r), e.g.:
> dbinom(x, size, prob, log = FALSE)
> pbinom(q, size, prob, lower.tail = TRUE, log.p = FALSE)
> qbinom(p, size, prob, lower.tail = TRUE, log.p = FALSE)
> rbinom(n, size, prob)
>
Sincerely,
Leonard
