@Antoni you shouldn't just agree that it's right because I think it is. I can be blind to my own errors sometimes.

I'd much rather you say there's a problem than let a possible error go

It's worth exploring a possible error

The reason I think it's right is because mu = n1 n2 / 2 and sigma^2 = n1 n2 (n1 + n2 + 1)/12

= [n1 n2 /2 ] x (n1 + n2 + 1)/6 = mu (n1 + n2 + 1)/6 .

But it's easily possible I'm overlooking something. It happens every day.

Statistics teaches us plenty about humility and our own limitations. Or at least it does for me.

(I even made a mistake typing that explanation up just now; since fixed, I think)

Hey buddy.

@Glen_b Hi Glen!

Thank you for explaining it!

Have a great day!

Busy procrastinating

I'm teaching a research subject to master students tomorrow. ... Brain not in gear yet

Hmm will this work? ¦̵̱ ̵̱ ̵̱ ̵̱ ̵̱(̢ ̡͇̅└͇̅┘͇̅ (▤8כ−◦

Does for me at least

.... .... ̵̄/͇̐\

Well, enough silliness, things to do

> We first used the Shapiro-Wilk formula to test the mentioned quantitative parameters for normality. In order to characterize the normally distributed variables, we determined their mean and standard deviation.

Is this a normal paragraph, or is the phrase "quantitative parameters" a bit strange? I'm a translator and am not sure yet whether the S-W formula can be used to test "parameters". I think it is used to test "samples".

any feedback on my Q (stats.stackexchange.com/questions/231086/…) would be much appreciated :) Cheers!

I'm just wondering if something is unclear, since it has gotten no answers and few comments, or could this be because it is not so trivial, maybe?

I'm in no hurry to get any response, but I'm curious about this, so don't want to see it disappear ;)

@Erosennin: Seems clear enough to me.

If you're using the full likelihood formulation (Poisson + offset) nothing will change - 5 counts in 10 hours is the same as 1 count in 6 hours & 4 counts in 4 hours.

If you're using the quasi-Poisson formulation you'll get the same point estimates but the standard error will change when your estimate of the dispersion parameter changes.

@Scortchi: Great! I'll look into the quasi-Poisson formulation vs the full likelihood formulation :)

Perhaps if you extend @HongOoi's examples you could answer yourself or at any rate fill out the q. a bit more to make it easier to answer.

I'll go in the direction you guided me, and try to extend his examples, in an attempt to answer it myself

@CowperKettle: You're right: it doesn't seem correct, even informally, to talk of testing parameters for normality.

@Scortchi Thank you! I will change it to "quantitative data"

I wonder what the difference is between "First-order carryover effects" and "second-order carryover effects" in a crossover clinical trial.

or "higher-order carryover effects" even

No. Not to "quantitative data".. hm.. "We first used the Shapiro-Wilk formula to test data for the mentioned quantitative parameters for normality. "

Awkward but better gets the original meaning across.

onlinecourses.science.psu.edu/stat509/node/126

> if first-order carryover effects are negligible, then higher-order carryover effects usually are negligible;

But what is the meaning of "first and higher order" here? O_O

@CowperKettle: 1st-order would be carry-over effects from treatments administered during period j - 1; 2nd-order from period j - 2; higher-order from all periods j > 1.

"Data for the mentioned quantitative parameters" reads oddly: just say "to test the data for normality".

@Scortchi Thank you, @Scortchi. I also managed to find a book with definitions on Google Books (0:

@CowperKettle: You're welcome.

@Scortchi Yes, indeed. But the Russian original text describes that data as "data corresponding to the mentioned parameters" (it does not mention the word "data" actually, but that is Russian technical lingo)

On the other hand, "mentioned data" might carry the same meaning in English in that context.

It's just an exercize. In a full-fledged translation I would have found a way to describe it.

In the curtailed context, it's harder.

Sure it's good Russian? By the way, it'd be "aforementioned data" - "mentioned" isn't used attributively in that way.

@Scortchi Thank you! Good point about "aforementioned", I'll try to remember that

Word of the day: Simpson's Paradox

@CowperKettle stats.stackexchange.com/search?q=Simpson%27s+paradox