Dobro jutro from Novi Sad

[jawnbc]

After a wicked long journey (22 hours door-to-door) today's my first real day in Serbia. Thanks to better living through chemistry and a decent night's sleep so far I've not fallen over in the street.
NS is a nice little city. There's an ubiquitous cafe culture and, as is often the case here in the former Yugoslavia lots of very tall 190cm men.
Jeebus took care of me during my talk this morning: a full room and only one quasi-asshat. Forage paper in first session--I'm a free man!

So far the calibre of the papers has been mixed. And I've forgotten what tools sone audience members can be. Oh and btw: if you haven't verified your sample is a Gaussian distribution...don't submit a paper with a t-test. I mean, come on. Really.

Posted via LiveJournal.app.

Comments

[snousle.livejournal.com]

if you haven't verified your sample is a Gaussian distribution...don't submit a paper with a t-test.

PET PEEVE TIME!!! Our pharma client actually demands a similar test on log- normal data. One little transform and everything would be fine. But they wrote it into their contract.

[jawnbc]

What is log normal data?

[snousle.livejournal.com]

Data that are normally distributed (gaussian) after you take the logarithm of each value. This is typical of gene expression measurements.

[jawnbc]

Ok....I'll take your word for it.

In this case it was a quasi-experiment involving two kinds of teaching methods. Social research is full of parametric stats with no evidence of a Gaussian distribution.

And don't get me started on sample size...

[jamaisneutral.livejournal.com]

oh, a t-test... brings back memories!

[jawnbc]

To t or not to to that is the question

[jeneration.livejournal.com]

But if the sample size is large, the t-test will still be accurate even if the distribution isn't gaussian, due to the central limits theorem - although I agree that at a minimum, the issue should at least be discussed!

[jawnbc]

I don't know of any social researchers using sample sizes in the thousands. Suffice to say there was no mention of a prospective power calculation either.