Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: R: RE: RE: InverseGamma with beta<0


From   "Carlo Lazzaro" <[email protected]>
To   <[email protected]>
Subject   st: R: RE: RE: InverseGamma with beta<0
Date   Mon, 18 Jun 2007 20:24:33 +0200

Dear Nick,

thanks a lot for Your Kind reply and for Your highly appreciated
teaching-notes.

As far as the replacement of SD with SE (as You correctly pointed out)in
Gamma distribution, it is suggested by the reference I quoted in my previous
message:

Briggs A, Schulper M, Claxton K. "Decision modelling for
health economic evaluation". Oxford: Oxford University Press, 
2006: 91-92.

I suppose that this replacement has to do with dealing with 
the uncertainty surrounding cost parameters (even though I don't know
whether this justification is sound enough from a statistician's point of
view).

I have generated random numbers following the intention to turn the base
case "static" cost model into a probabilistic one and see what happens, in
terms of uncertainty of cost at patient level, when this parameter is driven
by a stochastic factor (I mean random sampling from Gamma distribution with
an InvGamma distribution).

You are right once again about the confusion I have made about the term
InvGamma:

I actually meant the inverse of the (cumulative) distribution function 
of a gamma distribution, i.e. the quantile function 
of a gamma distribution, as yielded by beta * invgammap(alpha, probability).

 Thanks a lot one again for Your Kindness, time and hints.

King Regards,

Carlo

-----Messaggio originale-----
Da: [email protected]
[mailto:[email protected]] Per conto di Nick Cox
Inviato: luned� 18 giugno 2007 19.08
A: [email protected]
Oggetto: st: RE: RE: InverseGamma with beta<0

Thanks for this, which makes clearer what you did 
but leaves some questions on why you did it. 

Consider again this parameterisation for a two-parameter
gamma distribution: 

variable x >= 0 
shape parameter a > 0 
scale parameter b > 0

density function [1 / (b^a gamma(a))] x^(a - 1) exp(-x / b)

For this, standard results are that 

mean = a b 
variance = a b^2 




© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index