I am a new user to riskAMP and love it! I have the personal edition and want to link two (or more) distributions with a correlation coefficient. I can do this with Normal curves using "CorrelatedNormalValue", but I would like to link other distribution functions together like Weibull & Lognormal curves. Is there a way to correlate non Gaussian curves?

Hi,

Thanks for the nice words, but I'm afraid we don't support correlating non-normal distributions - at least not at the moment. I'll look into it and see if there's anything we can do. Are you particularly interested in the Weibull, or was that just an example? And more specifically, can you tell me anything else about what you're looking for?

Cheers,

Alex

Alex, for this spreadsheet I want to make a correlation between a curve fit of the historical consumer price index, and a second emperical distribution of future interest rates. (I am making the statement that there is a correlation between inflation rate and savings interest rates over time.) It is a spreadsheet I made to address the uncertainty of my retirement account. Inputs that I am modeling with distribution curves are; inflation rate, interest rate, medical inflation, beginning balance, and expenses. Output is the 10 percentile (90% chance of being greater) of my account value every year up until my wife is 94 (and I'm Kaput!). If I don't correlate the interest rate with the inflation rate, then during the simulation, individual calculations could occur with say; low inflation rate and high interest rate. that doesn't seem realistic to me.
At work I have Crystal Ball available to me, and so I did a fit for the CPI using it's curve fitting ability. The best fit (chi-square) for the CPI historical data is a logistical fctn. And I also like the Weibull fit. If I use a logistical fctn for the inflation rate and use "CorrelatedNormalValue" for the interest rate, I get a normal distibution for the interest rate. I would like to see something that looks like the Logistical fctn I am correlating to.
One way to "bypass" this would be to just add several percentage points to the inflation rate fctn and call that my interest rate, but I would like to be able to correlate the two with a correaltion value of less than 1. Brian

BLB350 Wrote:
-------------------------------------------------------
> Alex, for this spreadsheet I want to make a
> correlation between a curve fit of the historical
> consumer price index, and a second emperical
> distribution of future interest rates. (I am
> making the statement that there is a correlation
> between inflation rate and savings interest rates
> over time.) It is a spreadsheet I made to address
> the uncertainty of my retirement account. Inputs
> that I am modeling with distribution curves are;
> inflation rate, interest rate, medical inflation,
> beginning balance, and expenses. Output is the 10
> percentile (90% chance of being greater) of my
> account value every year up until my wife is 94
> (and I'm Kaput!). If I don't correlate the
> interest rate with the inflation rate, then during
> the simulation, individual calculations could
> occur with say; low inflation rate and high
> interest rate. that doesn't seem realistic to
> me.
> At work I have Crystal Ball available to me, and
> so I did a fit for the CPI using it's curve
> fitting ability. The best fit (chi-square) for
> the CPI historical data is a logistical fctn. And
> I also like the Weibull fit. If I use a
> logistical fctn for the inflation rate and use
> "CorrelatedNormalValue" for the interest rate, I
> get a normal distibution for the interest rate. I
> would like to see something that looks like the
> Logistical fctn I am correlating to.
> One way to "bypass" this would be to just add
> several percentage points to the inflation rate
> fctn and call that my interest rate, but I would
> like to be able to correlate the two with a
> correaltion value of less than 1. Brian