I dropped a coauthor who tortured data and ignored robustness checks.
VLRM asks me for data set: Da fuq I do?
-
Guy sent his data. I have no time to check, but he seems to have something. I still do not want him to fiddle with my data.
Claims, he gets same sign, better significance, slightly larger coef..
Apparently he did not (yet) do the robustness checks I'm afraid of.
I'm tenured and "friends" with the journal's editor. Worst thing that can happen to me for just ignoring him?This says a lot about the VLRM. If he was able to get the data that you yourself said took a substantial amount of time and money to get, the kid is obviously of high quality.
Either that or your RA was lazy.
-
This is all that is wrong about economics, though the trend is changing. JME, RED (tonname a few), have the code/dataset to replicate many of the published papers on their website. Some serious honest researchers (like ellen mcgrattan) have been putting the code and data up for replication for ages on their own websites. Any quant/empirical paper where the author does not make the data/code publicly available (or at least to the subscribers of the journal) or name sources where the data can be obtained, is not serious and is not science because it cannot be replicated. A serious system of peer reviewing would have referees being payed (and their name being associated with it) to reproduce the results, not just check for typos, whether the topic is trendy or if the paper cite the referee. See article in economist a few months back about the replicability of most papers in the medical sciences. Hell, I wrote some empirical/quant papers cherry picking some results and hiding others. Am I embaraced? yes. But then again, unless you are a star and you can make it anyway (david card is another example of honesty in that sense), you are hurt if you dont join the wagon. So OP, this is the price we all pay for not doing science. Take it like a man.
-
I actually made up most of my results up until now. Slight changes in order to fit my theoretical models.
Bad?Are you taking the proper precautions? Whenever I read about someone getting caught faking results, they didn't even take the time to make the distribution of first digits fit Benford's Law, even though that's the first thing I'd make sure to account for if I were making up data.
I wonder how many people out there are faking their results sufficiently well enough to never get caught.