.-
help for skewed bimodal generator
.-

Skewed bimodal generator
------------------------

         ^skbim^ p v1 v2 v3 v4 option [obs] [seednum] [sk1] [ku1] [sk2] [ku2]


Description
-----------

The program generates random numbers from a bimodal distribution. The two unimo
> dal distributions
that make up the bimodal can be normal or skewed-normal (see sknor for more det
> ails). Different
arguments can be inputted to the function, as specified by 'option'

p: the probability of the 1st unimodal distribution (obviously the prob for the
>  2nd will be 1-p)
option: takes values 1, 2 or 3 and defines v1-v4.
    for option=1: v1 the mean v2 the variance of the 1st unimodal, v3 the mean 
> v4 the variance of the 2nd unimodal
    for option=2: v1 the mean v2 the variance of the bimodal, v3 & v4 the varia
> nces of the unimodals
    for option=3: v1 the mean v2 the variance of the bimodal, v3 the mean v4 th
> e variance of the 1st unimodal
[obs]: number of observations to be generated (optional; set to current
    dataset size if omitted)
[seednum]: seed number to be used for drawing 'random' numbers - makes a differ
> ence if your simulations occur
    on the same second (i.e. running a batch of simulations)
[sk1]: skewness of the 1st unimodal (optional; set to 0 if omitted)
[ku1]: kurtosis of the 1st unimodal (optional; set to 3 if omitted)
[sk2]: skewness of the 2nd unimodal (optional; set to 0 if omitted)
[ku2]: kurtosis of the 2nd unimodal (optional; set to 3 if omitted)


Remarks
-------

When using options 2 and 3, parameters of the distribution may not be computed 
> for certain sets of
inputted values.
    for option 1 it must hold that: var - p*var(d1) - (1-p)*var(d2)>=0
    for option 2 it must hold that: var > p*var(d1) + p(1-p)*(m1-m2)^2
    where var the variance of the bimodal; m1,m2,v1,v2 means and variances of t
> he unimodals

For the default skewness and kurtosis values the generated distributions are no
> rmal.
The program is based on a method described in  "Ramberg et al, A probability Di
> stribution
and its uses in fitting data, Technometrics, 1979".
The method uses a formula with 4 parameters and parameter values which generate
data of various degrees of skewness and kurtosis have been provided in the pape
> r.
Only a few of those have been included in this program the command, but it can
easily be updated with more.
The included pairs of values are (sk,ku): (0,1.75), (0,3), (0,3.2), (0,3.3), (0
> ,3.4),
(0,3.6), (0,4), (0,4.4), (0,5), (0,6), (0,7), (0,8), (0,9), (0.5,3), (0.5,4), (
> 0.5,5),
(0.5,6), (0.5,7), (0.5,8), (0.5,9), (1,4), (1,5), (1,6), (1,7), (1,8), (1,9), (
> 1,10),
(1.5,6), (1.5,7), (1.5,8), (1.5,9), (1.5,10), (1.5,11), (1.5,12), (2,9), (2,10)
> , (2,11),
(2,12), (2,13), (2,14), (2,15),
For sk=2 and ku=9 the most extreme distribution is generated.
Negative skewness values can be inputted for left-skew distributions (the oppos
> ite values
of the ones listed above)


Examples
--------

 . ^skbim 0.5 5 1 7 1.5 1^
 . ^skbim 0.3 5 2 1.2 1.5 2 10000^
 . ^skbim 0.3 5 2 1.2 1.5 2 10000 1234^
 . ^skbim 0.5 9 3.5 8 2 3 10000 1234 1.5 6 -0.5 9^


Keywords
--------
bimodal, normal, distribution, skewness (skew, skewed), kurtosis


Author
-------

        Evangelos Kontopantelis
        National Primary Care Reserch and Development Centre
        University of Manchester
        M13 9PL
        e.kontopantelis@manchester.ac.uk


Also see
--------
    Ramberg et al paper: http://www.jstor.org/view/00401706/ap040083/04a00080/0
    STB:  STB-41 sg44.1, STB-28 sg44