Wednesday, April 19, 2017

This Is So Easy Even Someone Sick As A Dog Can See The Problem With It

I was asked to comment on a story in the news, that a couple of University of Kentucky Psychologists are claiming that the number of atheists in the United States, which other surveys have estimated at approximately three to ten percent* but which these guys are saying might be as high as 26%.  Well, instead of looking at Huffington Post or Democratic Underground, I looked at the preprint of the as yet unpublished article and the thing is so obviously bogus and so obviously designed to allow these people to make a big splash that anyone who takes it seriously only proves how credulous they are.

For example, take a gander at their "Genera Procedure" from which they "indirectly inferred atheism rates using unmatched count technique".

General Procedure
For both samples, we indirectly inferred atheism rates using the unmatched count
technique (e.g., Dalton, Wimbush, & Daily, 1994; Raghavarao & Federer, 1979),
a tool for inferring base rates of socially sensitive outcomes. The unmatched count technique indirectly infers underlying base rates for socially undesirable or unacceptable outcomes by randomly assigning participants to one of two versions of a count task. In one version participants indicate how many innocuous statements from a list (e.g., I can drive a motorcycle; I exercise regularly) are true of them. In the other version, participants receive a list that is identical, save for the addition of one sensitive item (e.g., I can drive a motorcycle; I exercise regularly; I smoke crack cocaine), and they indicate how many items are true of them. Crucially, nobody indicates which specific items are true of them, only how many in total. The difference between the aggregate rates in these conditions can presumably be attributed to the addition of the socially sensitive item. In using this task to indirectly measure atheist prevalence, our approach mirrors recent working using the unmatched count technique to indirectly estimate the size of the LGBT community as well as antigay sentiment (Coffman, Coffman, & Ericson, 2016). Crucially, this work includes extensive validation of the task’s utility in estimating the size of stigmatized groups, finding that the UCT does not appear to be driven by inattentive or random responding, and only generally diverges from self-reports of socially undesirable attributes (but not generic foil attributes). The task appears robust as well to participant inattentiveness and random responding (Coffman et al., 2016)

------
They claim their method "indirectly infers underlying base rates for socially undesirable or unacceptable outcomes by randomly assigning participants to one of two versions of a count task". Also note that they seem to be able to make it measure both the size of the LGBT community AND anti-gay sentiment.  Which would seem to me to be quite a feat, considering the entire things is based on inferences based in just what the researchers claim to want to see in their "data".  That is anything but an unproblematic or reliable procedure.

Do atheists consider their atheism socially undesirable or unacceptable?   I certainly have never gotten the impression that atheists in 2016 or even earlier feel that way.  What do the authors of the study base that in?  So, right off, even if you're credulous enough to take their methodology seriously - and I certainly don't - they might be basing their inference on a false assumption.  You couldn't possibly even begin to apply that without first finding out what percentage of atheists feel that their atheism is "socially undesireable or unacceptable".  Why not assume what you've found are members of religious minorities who might feel that atheists might be more approved of than they are.  After all, if the methods that measure atheists as 3 or even 10% of the population are unreliable, then those measuring social disapproval of atheism in the general population - what that often heard whine that atheists are a persecuted minority are based in - must be as unreliable.

My first question before reading the pre-print was why so many atheists were claimed to be such liars?  Now I know that they had nothing to base any of it on.

You wonder why atheists, after the past two decades of neo-atheist agitation, answering a poll in which their participation was anonymous would demonstrate a feeling that they would be shunned by answering the question directly.

If you want to press this there are other big, really big problems with the study which I could write about.  And since the news stories don't seem to even be based on the claims of the authors but in some vague feeling as to what they said, they're even more bogus.  They are click bait, nothing more.

All of this stuff, even those things that ask direct questions of opinions are based on the absurd faith that a. people respond honestly, b. people know the answer to what they are asked ,c. that if you asked them the same thing two weeks later that you would still get the same results and a myriad of other unfounded parts of the sociological faith system,  but none of them is reliable and none of them can really give you reliable results.  Social science (so called) has, after all reported absurd figures in which men claim to have had sex with more women than women have had sex with men and claimed that told you something about the population at large when that result is mathematically impossible. Yet they still report it and other "social scientists" whose training should have included a knowledge of mathematics sophisticated enough for them to know that result is impossible have based more "science" on such stuff.

It's all of entirely unknowable reliability.  For all any of us knows the actual figure is 33% or .3%.  It ain't rocket science, it ain't the far harder biological sciences - those which still are science, this is entirely outside of science which shows that merely pantomiming the methods of science doesn't produce science when the subject can't be honestly studied with them.

*  When I've looked at these things in the past, they ususally had to really twist things to come up with the higher percentages.

Update:  You have to wonder, if these two psychologists based at the University of Kentucky had claimed that atheists were only a percent or fewer of the population, what the articles at Huffington or Democratic Underground would be saying about it.  I suspect they would not have read the article but they would deride it as originating in Kentucky.  The authors would be trashed, made a joke.  You have to wonder what percentage of social sci-guys feel that such a study would be "socially sensitive" and what effect that would have on their "science".

No comments:

Post a Comment