Appropriate distribution for Rate in Probabilistic Sensitivity Analysis
Hi there,
I'm creating a Markov cohort model and have used Rate per person-years to calculate transition probabilities. I'm using the Rate to Prob function to convert the rate to a probability. Since my input in the model is rate, I would like to fit the distribution for a rate. I'm thinking of Gamma (not too sure) or Lognormal distribution which requires Mean and Median values to calculate mean of logs and std dev of logs.
The values I could source from literature was rate per person-year along with the confidence interval. What will be the appropriate distribution and 2) based on the suggested distribution, will I be able to calculate shape parameters from distribution parameters? All I have is the rate per person year and confidence intervals.
Looking forward to your help.
Regards,
Karan
Comments
I'm really not sure what distribution you would want to use for a rate. Perhaps someone else will have an opinion.
A decent option would be a Pert distribution. That would give you control over the upper/lower bounds as well as over variance. You could play with those parameters until you get a reasonable variance from the distribution that matches your confidence interval.
Andrew
Please sign in to leave a comment.