Confidence interval for data that cannot be lower than zero

[Houman]

Hi,

I'm trying to find the confidence interval for monthly investment account balances. I have calculated the mean and the standard deviation, however if I subtract 2xSD from the mean, the result is a a negative number. Given that the account holds long positions only, it's value cannot be negative. I'm wondering how would one then go about determining the confidence interval for such data. Just imposing a hard lower limit of zero while keeping the upper limit does not seem right. I'd appreciate any input on this.

Regards,
Houman

 

[tkhunny]

Re: Confidence interval for data that cannot be lower than z

The hard floor MAY be the right idea. More likely is honesty about your data. Should you be using a normal approximation? You may wish to onsider some sort of Gamma Dsistribution. You can look up other well-known distributions and see which meet you needs. If all you know is the Normal Distribution, this is likely to be a bit of a challenge. Where does that leave us?