I assume you have never had an Intro to Stats class. If you have, you didn't deserve to pass it. I suggest you look up a difference between two means test and categorical variables in regression.
I'm assuming here that you misunderstood this part. Though you quoted the entire post. Slashdot Beta?
Also, East-West bias can be noted in the stats measured and stats assumed.
No regression calculation was reported for West German family, while t-test values were always fixed (i.e. assumed) for East Germans and always calculated for West Germans.
And there's that thing of "East German family background" being marked with a 0 and "West German family background" being marked with a 1.
Someone seems to like West Germans better.
I am not talking about a single issue nor am I conflating their t-tests with their probit regressions.
I'm talking about several separate cases in the survey and in the paper where the language and variables used indicate either a pro-West or anti-East bias.
Which is basically code for "NOT-socialist" in this case, as seen below.
By the end of the paper they simply decide that doing calculations for "West German family" variable isn't needed.
"Meh, we found nothing there.
But look! We got this one point in a very small cherry picked sample that PROVES East Germans are cheaters!"
Nor am I misunderstanding surveying for dichotomous, binary, values.
I'm pointing out that the way the values are set up (i.e. a West German family which is a positive 1 and a NOT West German family - a 0) indicates a pro-West bias.
Which alone, doesn't mean much. But when you take in account the the rest of the paper...
Stuff like this:
As an interaction
term of age and family background is not informative in a non-linear model like Probit
we decided to investigate the exposure to socialism by
examining these distinct age cohorts separately.5 In line with the theory that exposure to
socialism impacts dishonesty, differences in cheating are greater in older cohorts. While East
German subjects born after socialism are 19 percent more likely to report the high side of the
die than their West German counterparts, subjects who lived less than 10 years in socialism
were 28 percent more likely, and subjects who lived for 20 years or more in socialism are 65
percent more likely.
I.e. "Because we can't find any proof for our original hypothesis, here's another cherry picked proof for an unrelated hypothesis - never mind the ignoratio elenchi taking place.
Wasn't it obvious we were gunning for a "proof" that socialism promotes cheating and not that a more general definition of East or West German family heritage does?
Why not just jump on another hypothesis as if we were proving that all the time, cause we can't find enough data to back up our claims for the original hypothesis? How about that?"
They had to get the sample down to 41 people (from a sample of 110, from a population of 259 - note how they only had 90 East Germans a table ago) in order to get ANY significant result regarding the age - cause that's what they are proving now.
Literally, that's the ONLY SIGNIFICANT RESULT IN THEIR ENTIRE REGRESSION CALCULATION.
That AGE of the subject matters and NOT his/her family heritage they've been talking about the whole time.
AND there is NO (zero, zip, nada, keine...) data presented for West Germans in that same period.
Apparently, there were no West Germans in 1970s but there were at least 41 East Germans.
That's cherry picking and replacing of the original argument with a different one, which though it sounds similar is ultimately irrelevant to the original hypothesis.
Plus they are misrepresenting results for other marginal effects of their probit regression - quoting them THOUGH they are not statistically significant for the p-value they've set up.
The link to the paper is in the summary. Go look it up.