How to put distribution in denominator?
Sibo Liu
Dear Community,
I'm wondering if someone have used a distribution in the denominator?
I am doing a probability sensitivity analysis, and created a normal distribution of median survival time with mean and SE, and plan to draw random number from this distribution for each monte carlo simulation. I used the exponential distribution to model patient survival, and thus the lambda for the exponential distribution is ln(2)/the number each draw from the normal distribution of median survival.
But I encountered an issue, when putting the distribution in the denominator, it warns me 'Division by zero' and the PSA cannot run. Can someone kindly comment on this if you have experienced that before please? Thank you so much!
Sibo
Comments
The distribution you are using in the denominator presumably draws a value which can be zero at some point.
You can definitely use a distribution in the denominator - just suggest you Graph the samples which are draw from the distribution and ensure none are zero.
Just to follow up on that, is there a place that I can set the upper / lower limit of the normal distribution, or I should use some function to make the drawed value to be non-zero please? Thanks
You cannot bound a normal distribution. It has long tales in both directions.
You could try a Pert distribution, which will have fixed bounds on the low/high end of the range. Use the CDF/PDF graphs to the right of the distribution to see how the distribution behaves as you change the parameters.
Please sign in to leave a comment.