On Sun, 28 Aug 2016 20:24:04 -0700, Mike Palij wrote:
Although I have been somewhat following the "replication crisis"
I did not know the "sad face" result and its apparent importance.
Failure to replicate this result is the focus of this article in Slate;
see:
http://www.slate.com/articles/health_and_science/cover_story/2016/08/can_smiling_make_you_happier_maybe_maybe_not_we_have_no_idea.html
In looking over the Slate article above, it occurred to me that
the author and some of the researchers he quotes misses
some important points, namely:

(1) Psychologists do experimental research in order to establish
causal relationships between/among variables.  What make
scientific research valuable is that these causal relationships
should hold up either in all conditions (i.e., a main effect; such
as gravity operates the same throughout the universe) or
interacts with other variables such that the causal relationship
changes in magnitude or even presence (i.e., disappears;
antibiotic treatment of tuberculosis is more effective with
inpaitent groups who receive the treatment on a regular schedule
as well as other supports as compared to a homeless person
with HIV/AIDS who has to self-administer meds). The
establishment of a causal relationship is a critically important
part of our valid knowledge of reality.

(2) Issues regarding whether or not a causal relationship exists
fall under the heading of internal validity and one does not have
to read Campbell & Stanley every night before going to sleep
(or does one?) to be reminded of the threats to internal validity
as well as factors that can give rise to spuriously significant
result (i.e., Type I errors, experimenter effects and other reactive
effects).  The Slate article does not mention the importance
of establishing causal effects in experiments, the degree of
internal validity, and the threats to internal validity (to be fair,
there are a few comments that can be interpreted as threats
to internal validity but these are presented without the validity
context).

(3) The extent to which a causal relationship exists beyond
the original situation (i.e., the experimental situation) is an
issue concerning external validity and ecological validity.
Treatment of tuberculosis is an example of both:  (a) some
antibiotics may or may not work with certain strains of TB,
and (b) the schedule of everyday treatment should follow as
closely as possible the schedule used in the experimental
study -- something that is reasonable in an inpatient treatment
facility or with people with stable living conditions with good
social supports but unreasonable for homeless people with
other serious illness.  I raise this point because at the end of the
article Strack, defender of the "sad face" effect, comments
on a meta-analysis that found that roughly half of the replication
found the effect and half did not, overall no causal relationship.
But the article says the following:

|So when Strack looks at the recent data he sees not a
|total failure but a set of mixed results. Nine labs found the
|pen-in-mouth effect going in the right direction. Eight labs
|found the opposite.

So, toss a coin and you'll get the profanity of getting
a statistically significant or nonsignificant result (or, for
people with NHSTphobia, an effect size greater than
zero or an effect size equal to zero or less than zero).

||Instead of averaging these together to get a zero effect,
|why not try to figure out how the two groups might have
|differed? Maybe there's a reason why half the labs could
|not elicit the effect.

Maybe, but the most parsimonious explanation is that there
is no effect.  Strack appears to be making an argument
based on the limited generality of the causal relationship,
that it interacts with other variables, and this is why it is
present in some research but absent in others.  But this
is a post hoc explanation and the Slate article does not
identify which variables might be operating to either
suppress the causal effect or to allow it to be expressed
(to be fair, I have not read Strack's original article on this
point and I do not know if he has shown that re-analyses
of data from these studies show which variables
suppress/express the effect, then he may be on solid ground
though one might argue that he is on a fishing expedition
in such analyses -- if certain 3rd variables keep showing up,
then one has to do the experiment that has conditions with
that/those 3rd variables present and absent, and showing
that the effect depends upon the presence/absence of these
variables, otherwise he's just blowing smoke).

Again, I am reminded of the DiCara situation and Miller
& Dworkin's inability to replicate DiCara's original finding
of conditioning of the autonomic nervous system: M&D
still believe the effect was real even though they were
not able to replicate it -- nor has anyone else since the
publication of their results.

Perhaps researchers need to be reminded of the tentativeness
of our knowledge and better appreciate which causal
relationships can be reliably shown and which cannot.

-Mike Palij
New York University
[email protected]





---
You are currently subscribed to tips as: [email protected].
To unsubscribe click here: 
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=49349
or send a blank email to 
leave-49349-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu

Reply via email to