Sandbox posts last active a week ago: Test cases

let's maybe not star that every time it happens

considering it's already happened twice

and instead encourage @AZTECCO to use h1s for his sandbox titles :P

@UnrelatedString fixed (closes issue #62446770)

lmao

Hello!

@RadvylfPrograms That could be really cool

Does JS let you programmatically get the stuff inside a label?

Because then you could use JS labels for assembly labels

Unfortunately it does not

:(

However, you could use an object literal, since they are not lazily evaluated and their keys are ordered

There go my dreams of making the competitor to wasm :P

Ooh nice

Not sure how you'd do .text and stuff

I bet this'd be doable in Groovy

Syntax still need some work, but a start

Worth noting it doesn't matter what you actually do with the identifiers since as soon as they're evaluated the work is done. Maybe something like mov- rbx, rsi would look nicer

Basic stats question. I have 100,000 items each of which is in one of five categories and each item has a price. I want to estimate the proportion of cost for each category. If I sample 25 at random and just use the price of the items in that sample, how should do the random sample?

Should the probabilities be weighted by the price of each item?

You're overthinking it, if you have N items from a category in your random sample, the expected price of that category as a fraction of the total is just the sum of prices of those N items divided by the total in your sample

You could also approach it as the expected average cost of an item in that category multiplied by the expected percentage of the items that are in that category, which should give you the exact same result

@RadvylfPrograms I was thinking that if I have one item with cost 1000 and one hundred with cost 1, a uniform random sample will get it wrong

That's just a limitation of small sample sizes and statistics in general

@RadvylfPrograms but if weight my sample by cost, don’t I get it right?

No matter how you sample your data or account for things, it's always possible there's some wild outlier that makes your predictions totally inaccurate

@graffe Oh as in, sampling non-uniformly?

@RadvylfPrograms yes

I guess it'd depend on how much information you have. Like, at some point you need enough information to be able to sample well that you could just use a better approach than random sampling, or more than 25 samples

I have all the prices in an array so weighted sampling is easy

Determining which category an item is is done by hand so is expensive

Oh interesting

Thanks :)

it’s actually a practical problem

I guess if there is a large disparity in prices between things and you need the minimum possible number of samples, you'd probably want the number of samples in a subgroup of items to be roughly proportional to the sum of their costs

Yes

E.g. if you have one item that costs $1000 and 400 that cost $100 each you'd want 20 and 5

But if I weight my sampling by the cost of each item, is my estimate “unbiased”?

Oh that's a good question

I think so

Thank you! I am much better at questions than answers :)

Also, do I want to sample without replacement? That is never sample the same item twice?

Probably

This is a little out of my expertise though so I'm just going by intuition

Mine too!

Thank you though

I don't think it would bias the result either way, but with replacement seems like it'd be something you'd only do if you know you should

@RadvylfPrograms yes I was hoping someone could tell me :)

good morning!

quiet time here?

very quiet!

From the chat rules: "Sometimes, there isn't anybody talking in chat. That's perfectly fine. Don't send messages just because the room is quiet." See here: cgcc-se.github.io/chatiquette

:(

Hiya

How come ruby interprets // as regex even though it isnt lol

because // is an empty regex

if there's stuff between the // that gets treated as the pattern

rubyguides.com/2015/06/ruby-regex

the empty regex can be useful in golfing, especially in perl

@Ginger new PFP?

exactly

and if I don't finish before December Niko gets a santa hat

also, we're at 23 questions for CGAC and we only have 4 days until December!

:o

@Ginger oh no

good afternoon, just wanted to say hi and share this: loda-lang.org - it might be interesting for oeis enthusiasts!