Ten fold

7:23 AM

@gung Interesting. Are "4 SO sites" just SO in different languages? I didn't even know there was SO in more than English and Russian... Can you post a link to the stats for sites?

amoeba

8:25 AM

@StephanKolassa stackexchange.com/sites?view=list#questionsperday

alhelal

9:27 AM

Why the name solve() for invert matrix in R?

Mario Migliaccio

11:56 AM

Hi, do you anyone or any site that provides tutoring in time series analysis with reasonable prices?

gung

12:40 PM

@amoeba, I believe the technically correct usage would be to say: N(0, 1) is a distribution, N(100, 15) is a distribution, N(mu, sigma^2) is a family of distributions.

@MarioMigliaccio, I don't, but there are free TS tutorials and books available on the internet, even pretty good & comprehensive ones. You could ask specific questions on CV when you get stuck.

whuber

3:44 PM

@alhelal Matrix inversion is equivalent to solving a system of simultaneous linear equations--and indeed, that's how it's often carried out in numerical algorithms.

@amoeba (cc @gung) Gung's right, but that still doesn't prevent (many) people from committing such abuses of language. It's common to read passages that call N(mu,sigma^2) "the normal distribution," for example. The important word "family" that should follow this phrase is omitted. Thus you have to read with caution and always be prepared to infer the meaning from the context.

@amoeba Your uses of "parameters" and "family" are unusual, though. For instance, the exponential family is a one-parameter family of distributions: it is indexed by a single positive number.

Mario Migliaccio

4:21 PM

@gung Well I did, but I'm still got stuck, I don't know if I should repost my question in a different manner

amoeba

@whuber But there are several continuous functions in the definition of exponential family! So exponential family is a family (set) of sub-families (sets) of distributions, where each distribution inside the sub-family is indexed by a single number, but sub-families are indexed by several functions... This is cumbersome, no wonder that people simplify

gung

4:41 PM

@whuber, @amoeba, I'll admit that I often abuse the terms in that way. I think it can often be preferable in context to speak in a manner that is not the most technically correct.

@MarioMigliaccio, is the problem that you've asked some question & no one is answering? I'm not sure I follow you.

ALannister

6:19 PM

0

Q: Induced distribution measure and induced distribution function where original r.v. is Pareto

Consider for one car owner the insurance policy with the following clauses: Deductible: If the loss $X>d$, then the insurer pays only for loss above $d>0$. Coverage Limit: If the loss $X>l$, then the insurer pays only for loss below $l>d$. Distorted Distribution: The insurer may base the premiu...

I need help with this question very badly. If there is anyone around who can help, I would be forever grateful! It's not getting a lot of attention on the main page.

Kodiologist

6:34 PM

@ALannister As of this writing, that question is 40 minutes old. Response times are rarely that quick on this site, especially for such a long question.

ALannister

That's good to know. I need to figure it out today, though. Because I need to have this part figured out in order to go on to the next part.

Kodiologist

@ALannister You can try bountying it, but there still isn't much of a guarantee of getting a reply within 12 hours.

ALannister

Can't bounty it until it's 2 days old.

Kodiologist

Oh, right.

ALannister

I also have this question listed on MSE. There it has a 100 point bounty, but I've alienated all the probability people over there.

I just found out this place existed.

Kodiologist

6:42 PM

You know you're not supposed to cross-post, right? I think it's fine to advertise your math.SE question in this chat, but not to repost it in CV without deleting the math.SE version.

ALannister

No, I didn't know that.

Oh crap.

Kodiologist

2

Q: Induced distribution measure and induced distribution function where original r.v. is Pareto

Consider for one car owner the insurance policy with the following clauses: Deductible: If the loss $X>d$, then the insurer pays only for loss above $d>0$. Coverage Limit: If the loss $X>l$, then the insurer pays only for loss below $l>d$. Distorted Distribution: The insurer may base the premiu...

probability integration probability-theory probability-distributions

ALannister

Yeah, that's it.

Kodiologist

It's okay; you can just delete the CV version, and there's your link.

ALannister

All right. :( I hope I get an answer. There's a lot hanging in the balance of whether I can figure this out.

It's deleted here.

Nick Cox

6:51 PM

@ALannister Perhaps you need to reinvent yourself as a Stark.

ALannister

Haha. The North remembers, I know.

@NickCox seriously though, can you help me?

Nick Cox

Would if I could. Way off my beaten track. Sorry.

whuber

7:27 PM

@amoeba I don't understand your remarks. The Exponential distribution family consists of distribution functions F(x) = (1 - exp(-k*x))*I(x > 0) for positive real values $k$. That's it--one parameter.

DeltaIV

7:40 PM

Hi, probably a very stupid question but if @whuber, @GavinSimpson or anyone else knowledgeable in GAMs would like to have a look at it...

0

Q: Are GAM models linear in the parameters?

Consider a GAM model, expressed in mgcv just to fix ideas: my_model <- gam(y ~ ti(x1)+ti(x2) + ti(x1, x2), method= "REML") The model is linear in the parameters, right? Each smooth is a linear combination of basis functions, which are independent of the data set (unless I use bs = "ad"). Thus ...

linear-model gam mgcv

amoeba

@whuber OK, I understand the confusion. I was talking about en.wikipedia.org/wiki/Exponential_family. This is a nice example because it exemplifies the potentially ambiguous terminology. You were talking about what is usually called "exponential distribution" but is actually a family; however it's not the same family as what is usually called "exponential family"

whuber

10:08 PM

@amoeba Yes, it's a good example of ambiguity. I don't think it exemplifies the issue we have been discussing, though, because "an exponential family" is not used in the sense of hypothesizing a set of distributions in a model, nor would one ordinarily think of exponential families being "parameterized" by functions (continuous or not).

Transcript for