On the plus side it’s forcing people to acknowledge the uncertainty involved in many numbers they use.
On the minus side it’s treating everything as a normal (Gaussian) distribution. That’s a common default assumption, but it’s not necessarily a good assumption. To start with an obvious problem, a lot of real-world values are bounded, but the normal distribution is not.
It’s open source. Right now I only know very basic Python, but I’m taking a CS course this coming semester and I’m going for a minor in CS. How hard do you think it would be to add in other distributions, bounded values, etc.?
As a matter of programming it would be very easy. The difficult part is designing the user interface so that the availability of the options doesn’t make the overall product worse.
Author is on the effective altruism forum, he said his next planned future is more distributions, and that he specifically architected it to be easy to add new distributions.
How hard will it be to add features depends on the way it’s architected, but the real issue is complexity. After you add other distributions, bounds, etc. the user would have to figure out what are the right choices for his specific situation and that’s a set of non-trivial decisions.
Besides, one of the reasons people like normal distributions is that they are nicely tractable. If you want to, say, add two it’s easy to do. But once you go to even slightly complicated things like truncated normals, a lot of operations do not have analytical solutions and you need to do stuff numerically and that becomes… complex and slow.
Link: Introducing Guesstimate, a Spreadsheet for Things That Aren’t Certain
How useful do you think this actually is?
Moderately.
On the plus side it’s forcing people to acknowledge the uncertainty involved in many numbers they use.
On the minus side it’s treating everything as a normal (Gaussian) distribution. That’s a common default assumption, but it’s not necessarily a good assumption. To start with an obvious problem, a lot of real-world values are bounded, but the normal distribution is not.
It’s open source. Right now I only know very basic Python, but I’m taking a CS course this coming semester and I’m going for a minor in CS. How hard do you think it would be to add in other distributions, bounded values, etc.?
As a matter of programming it would be very easy. The difficult part is designing the user interface so that the availability of the options doesn’t make the overall product worse.
Author is on the effective altruism forum, he said his next planned future is more distributions, and that he specifically architected it to be easy to add new distributions.
How hard will it be to add features depends on the way it’s architected, but the real issue is complexity. After you add other distributions, bounds, etc. the user would have to figure out what are the right choices for his specific situation and that’s a set of non-trivial decisions.
Besides, one of the reasons people like normal distributions is that they are nicely tractable. If you want to, say, add two it’s easy to do. But once you go to even slightly complicated things like truncated normals, a lot of operations do not have analytical solutions and you need to do stuff numerically and that becomes… complex and slow.
It is already doing everything numerically.
This is awesome. Awesome awesome awesome. I have been trying to code something like this for a long time but I’ve never got the hang of UI design.