ASA Connect

I am preparing a revenue forecast for a company that caters special events like weddings. I would like to use a simulation package that offers a number of continuous and discrete distribution options. Total revenue for the year is based on the number of events catered x the number of guests x per guest fee.

So, we have two discrete variables and one continuous one. Unfortunately, historical data is sparse so I do not have the ability to develop an empirical distribution for any of these variables. What I do have for each variable is a good fix on the minimum, maximum and a management opinion on a modal value.

If all the variables were continuous, I would have little qualms about choosing either a Triangular or PERT distribution, both of which are of course continuous models. But I am wondering if utilizing either of these distributions to model the discrete variables will lead to any systematic biasing of the forecast.

The discrete distributions offered in the package such as Poisson will not work. I have been searching for the last day literature on this topic but have been unable to locate any discussions.

I realize that there are precedents for using continuous distributions to approximate discrete distributions (using the Normal to approximate a Binomial). But this latter example also specifies a correction factor. The only correction factor I can think of if I apply a Triangular or PERT distribution to the number of events or number of guests would involve using @Round to the simulation outputs. Hardly a solution based on any theory. I would appreciate any suggestions on how to proceed.

I am preparing a revenue forecast for a company that caters special events like weddings. I would like to use a simulation package that offers a number of continuous and discrete distribution options. Total revenue for the year is based on the number of events catered x the number of guests x per guest fee.

So, we have two discrete variables and one continuous one. Unfortunately, historical data is sparse so I do not have the ability to develop an empirical distribution for any of these variables. What I do have for each variable is a good fix on the minimum, maximum and a management opinion on a modal value.

If all the variables were continuous, I would have little qualms about choosing either a Triangular or PERT distribution, both of which are of course continuous models. But I am wondering if utilizing either of these distributions to model the discrete variables will lead to any systematic biasing of the forecast.

The discrete distributions offered in the package such as Poisson will not work. I have been searching for the last day literature on this topic but have been unable to locate any discussions.

I realize that there are precedents for using continuous distributions to approximate discrete distributions (using the Normal to approximate a Binomial). But this latter example also specifies a correction factor. The only correction factor I can think of if I apply a Triangular or PERT distribution to the number of events or number of guests would involve using @Round to the simulation outputs. Hardly a solution based on any theory. I would appreciate any suggestions on how to proceed.

I am preparing a revenue forecast for a company that caters special events like weddings. I would like to use a simulation package that offers a number of continuous and discrete distribution options. Total revenue for the year is based on the number of events catered x the number of guests x per guest fee.

So, we have two discrete variables and one continuous one. Unfortunately, historical data is sparse so I do not have the ability to develop an empirical distribution for any of these variables. What I do have for each variable is a good fix on the minimum, maximum and a management opinion on a modal value.

If all the variables were continuous, I would have little qualms about choosing either a Triangular or PERT distribution, both of which are of course continuous models. But I am wondering if utilizing either of these distributions to model the discrete variables will lead to any systematic biasing of the forecast.

The discrete distributions offered in the package such as Poisson will not work. I have been searching for the last day literature on this topic but have been unable to locate any discussions.

I realize that there are precedents for using continuous distributions to approximate discrete distributions (using the Normal to approximate a Binomial). But this latter example also specifies a correction factor. The only correction factor I can think of if I apply a Triangular or PERT distribution to the number of events or number of guests would involve using @Round to the simulation outputs. Hardly a solution based on any theory. I would appreciate any suggestions on how to proceed.

I am preparing a revenue forecast for a company that caters special events like weddings. I would like to use a simulation package that offers a number of continuous and discrete distribution options. Total revenue for the year is based on the number of events catered x the number of guests x per guest fee.

So, we have two discrete variables and one continuous one. Unfortunately, historical data is sparse so I do not have the ability to develop an empirical distribution for any of these variables. What I do have for each variable is a good fix on the minimum, maximum and a management opinion on a modal value.

If all the variables were continuous, I would have little qualms about choosing either a Triangular or PERT distribution, both of which are of course continuous models. But I am wondering if utilizing either of these distributions to model the discrete variables will lead to any systematic biasing of the forecast.

The discrete distributions offered in the package such as Poisson will not work. I have been searching for the last day literature on this topic but have been unable to locate any discussions.

I realize that there are precedents for using continuous distributions to approximate discrete distributions (using the Normal to approximate a Binomial). But this latter example also specifies a correction factor. The only correction factor I can think of if I apply a Triangular or PERT distribution to the number of events or number of guests would involve using @Round to the simulation outputs. Hardly a solution based on any theory. I would appreciate any suggestions on how to proceed.