Noob question about 32-bit ints

[alexhallam]

I was reading through the users manual and came across this.

Stan uses 32-bit (4-byte) integers for all of its integer representations. The maximum
value that can be represented as an integer is 2^31 − 1

so 2^31 -1 is about 2.9 billion. So, if your outcome variable was company dollars that were over 2 billion dollars then you could not use Stan?

This does not seem right.

[sakrejda]

@alexhallam

ugh the users manual and came across this.

Stan uses 32-bit (4-byte) integers for all of its integer representations. The maximum

value that can be represented as an integer is 2^31 − 1

so 2^31 -1 is about 2.9 billion. So, if your outcome variable was company dollars that were over 2 billion dollars then you could not use Stan?

With amounts like that just use doubles.

[alexhallam]

Thats what I figured. I just wanted to make sure that the limit was that low.

[sakrejda]

Yeah this is an issue everywhere, for us it’s been difficult to deal with in generated quantities where sometimes an RNG can produce a number too large to represent and then throw/crash the model. I think we did something about that. I think R only recently went to 64-bit integers everywhere (?) but yeah, main question to ask is whether you really need exact numbers.

[bgoodri]

We are hopefully going to move to 64-bit ints soon

Issue: Widen int to 64-bit signed integer in generated code

but yeah, you should just use real numbers for dollar amounts.