itt get the mafia statistics you crave

[Iecerint]

No, I think he means like "how many times Iecerint has been in lylo divided by Iecerint's total mafia games" and similarly for each player. So each bar is a number of times a certain player has been in lylo adjusted by the number of games.

Lotsa bars. You could select only players with at least N games to make it more manageable.

[Iecerint]

It might be more interesting to make it "all lylos" instead of "all 3p lylos" IMO.

[Psyche]

logistic regression project wins jsut because i already have most of the data

[Iecerint]

Make sure the

p

value is low enough to publish, first.

[Psyche]

do you not want to see it if the p value is high

[Empking]

Ircerint's totally corrupt. Look at this, and the p-value's super high *hides stacks of papers behind his back*

[N]

Are you still taking requests? I was going to make a list of roles used in Normal games, but then I realised maybe you have an easy way to figure that shit out!

[wgeurts]

They're all on the wiki. Just look for normal game or something.

[N]

Not roles allowed in a Normal game; roles used in Normal games.

[Iecerint]

do you not want to see it if the p value is high

If the

p

value is high, then it doesn't really say anything. For example, it could be that your logistic regression model is incorrect instead of that the effect does not exist (i.e., since this will inflate the standard error of the relevant term).

It should be a fun exercise either way.

[Psyche]

Are you still taking requests? I was going to make a list of roles used in Normal games, but then I realised maybe you have an easy way to figure that shit out!

I built this program a long time ago that looked for a playerlist with roles in a mod's iso, but since mods are imperfect so is the program. In general, this is pretty hard to get right!

[Psyche]

do you not want to see it if the p value is high

If the

p

value is high, then it doesn't really say anything. For example, it could be that your logistic regression model is incorrect instead of that the effect does not exist (i.e., since this will inflate the standard error of the relevant term).

It should be a fun exercise either way.

i really need to review my statistics!

[Empking]

Are you still taking requests? I was going to make a list of roles used in Normal games, but then I realised maybe you have an easy way to figure that shit out!

I built this program a long time ago that looked for a playerlist with roles in a mod's iso, but since mods are imperfect so is the program. In general, this is pretty hard to get right!

If you only care about Normals. Then you can just use the mini normal archive for most of it.

[Psyche]

oh well if i can just do that it's easy and ive done it before

[Iecerint]

do you not want to see it if the p value is high

If the

p

value is high, then it doesn't really say anything. For example, it could be that your logistic regression model is incorrect instead of that the effect does not exist (i.e., since this will inflate the standard error of the relevant term).

It should be a fun exercise either way.

i really need to review my statistics!

A high

p

value is just failure to disprove the null. It doesn't prove the null.

That's why publishing negative results is really hard. You can only do it if you show that you really went overboard attempting to control or account for every possible moderating factor and you had incredible power, etc.

[Psyche]

oh i know all that
i thought you were suggesting that i could get a low p value in support of the null hypothesis
given everything i've done on activity, I'm pretty confident that getting a low p value for the thing you want me to do is going to be hard

[Iecerint]

What format is your data in?

[Iecerint]

given everything i've done on X, I'm pretty confident that getting a low p value for X is going to be hard

This is how I feel about gender effects on brain activity btw, but people somehow manage to make a career out of it.

[House]

Word(s) players tend to mirror from game to game as vanilla/PR (per each PR)/scum

Wagon placement comparison as vanilla/PR/scum

Player's tendency to bus during RVS (in %)

Player's tendency to bus (separate %)

[PeregrineV]

% chance of scum in each position of day1 mislynch for Normal 13p game, normal 21p game, Normal 24p game

% of time scum is in first 50% of votes on lynched teammate.
% of time scum is in last 50% of votes on lynched teammate.

[Iecerint]

given everything i've done on X, I'm pretty confident that getting a low p value for X is going to be hard

This is how I feel about gender effects on brain activity btw, but people somehow manage to make a career out of it.

Don't you have like thousands of degrees of freedom? It should be easy to get a low

p

value for any trivial question you would ask.

[Psyche]

i need to do quite a bit of manual work to get it that high
but i guess i will

[choof]

if you do a visual representation of your own post history is that considered self-meta analysis