Originally Posted by
exerda
To be fair, with self-reporting from FTers, there's going to be a selection bias present in the data regardless of how large the data set is--unless as you say it covered all pax and flights.
There is a major bias, but its not a
selection bias, which refers to bias in the subjects that are chosen by a studies authors. Here that has not been done as no sample was selected or pulled by the authors.
What there is a "self-selecting bias" which refers to those participating
selecting themselves. Polls done this way (like those great pop up polls on Fox news, etc "is Obama a communist") are called "SLOPs"
A good basic discussion of these type of studies (and their lack of statistical validity) is here:
http://www.aapor.org/Content/aapor/R...es/default.htm
Originally Posted by
mgcsinc
No doubt about that.
What bothers me, though, is that some people apparently come to the table with the notion that selection bias = completely unusable results.
The key is to think critically about the selection bias and what it's likely to do to the data. Take brain imaging studies, for example. Most subjects in those studies are 18-26 year old smart people. (I'll let you figure out why.) That's some pretty big selection bias. And yet, we accept it and move on, because the bias should have no predicted effect on the research questions being studied. To the extent that there are concerns, these are noted and recognized as limitations of the study.
My hope is that people here can move beyond their initial skepticism and accept the data on DidItClear while recognizing that the limitations on the data are themselves limited.
I post my own Data on DidItClear. I find the results interesting.
I am glad that you (and others?) put the site together.
I thank you for your work. But don't for a moment think that the results are statistically valid for any purpose, they are not. Once you rely upon people to select themselves you are no longer representative period. What you have is the results of those who felt strongly enough about their experiences to post. Its not a statistical sample at all. What you may have (and its interesting) is some information on when upgrades are most/least likely to clear, and at what cities. [This is assuming people input all of their data, not selected data
]. Again its interesting data, but there is not a sample, let alone a random sample, and you can't extrapolate from it under well established rules of statistics.
Frankly, the data that you would get from asking 10 random folks in Group 1/2 at the airport waiting to board a UA flight would be better (as you have at least pulled a sample, if not a random one). Likewise, I find the upgrade percentages that people post on FT to be interesting (note I posted my 2013 upgrade rates in this thread) but again it has a clear self selection bias to it.
Originally Posted by
LarkSFO
The same bias is, of course, present here in FT reporting in this thread...
UDU stats and now did it clear at least capture the information in a somewhat organized fashion.
But yes, any self reporting system will have bias.
Yup.
Originally Posted by
Chris1000
Yield management?
Sale?
Discounting something that has not been bought at a higher price is not a reverse auction.
I appreciate the suggestions, but I don't think that any of these things capture the essence of what I was describing as well, and I think people understood what I was saying/getting at.
