I’m attempting to use Causal to go through the Applied Information Economics process, as described in in the Book “How to Measure Anything”.
Causal is great at the monte-carlo modelling part, but there’s a step after that to calculate the value of perfect information for each variable, which shows you which subsequent variables it’s worth measuring.
Here’s how it’s described in the book.
- Create a loss function that shows how much you lose over the range of values.
1.Slice the distribution of the variable you’re wanting to measure into hundreds or thousands of small segments. In the case of the units sold, I might want to create a spreadsheet where I show a table with increments of 1,000 units. Each row shown is actually the middle of a small range starting at 0 units sold all the way up to 200,000 units sold. I don’t need to go any higher because results greater than 200,000 units would have a loss of zero. (If you do go past 200,000 units, be sure to add an if() statement to make the loss zero if units sold is greater than 200,000.) That would make a table of 200 rows.
- Use the loss function to compute the opportunity loss for each segment. For each row in the table, use the loss function to compute a loss. For the row with 100,000 units sold, our loss would be (200,000-100,000)*25=$2.5 million . . . and so on for all the other rows.
- Compute the probability for each segment. For normal distributions, use the method shown in the “Computing an ‘Incremental Probability’” inset (it uses the normdist() function in excel).
- Multiply the opportunity loss of each segment by its probability.
- Total all the products from Step 4 for all segments.
Hubbard, Douglas W… How to Measure Anything: Finding the Value of Intangibles in Business (p. 151). Wiley. Kindle Edition.
Any idea of how to do this sort of discrete approximation using Causal?