I forgot to say that one should use "open label" as the keyword for searching PsycInfo for the drug studies when participants and researchers know what drug is being used.
-Mike Palij New York University [email protected] ---------- Forwarded message ---------- From: Michael Palij <[email protected]> Date: Tue, Mar 27, 2012 at 6:54 PM Subject: re: [tips] Examples of poorly designed studies? To: "Teaching in the Psychological Sciences (TIPS)" <[email protected]> Cc: Michael Palij <[email protected]> On Tue, 27 Mar 2012 14:37:24 -0700, Jim Clark wrote: >Hi > >I've been asked to cover the last few classes in Research Methods, >which I have never taught (or at least not in living memory). I wonder >if people have examples of poorly designed studies that lend themselves >to brief descriptions? I don't present poorly designed studies as examples in class so I don't have any specific studies in mind but, in keeping with the recent discussion on expectancy effects, a search of PsycInfo turns up hundreds of studies where both patients and researchers knew that a drug was being administered. This sets up a discussion why (a) the result has to be replicated before recommended for use and (b) why a double-blind placebo control design has to be used. NOTE #1: You can contrast drug effects with other effects in psychology, such as those shown by the implicit association effect where what you know may not affect the outcome (as trying to game the experiment is detected by an increase in reaction time showing a shift from automatic processing to controlled processing). For example, see this video from Scientific American Frontiers where Mahzarin Banarji and Brian Nosek can't help but reveal the pattern of association among their social concepts: http://www.youtube.com/watch?v=2RSVz6VEybk Note #2: You can show the same effect with a levels of processing set of instructions under appropriate conditions. The http://opl.apa.org website's self-reference experiment demonstrates this though I'm not sure they fixed up the labeling problem in output. Then again, you can have students read Milgram's "Behavioral Study of Obedience" and have them identify (a) the independent variable and (b) the control or reference condition (against which the "treatment" condition is compared). It become apparent there is no independent variable (there is only one level that subjects experience -- one has to read Obedience to realize that the several studies varying distance between "teacher" and "learner" actually constitute one experiment, albeit conducted over time). There, of course, is no real control condition, just the informal questioning that Milgram engaged in before the conduct of the study (i.e., asking people if they would go "all the way"). Similarly, one could do the same thing with the Zimbardo Stanford Prison experiment -- there is only one level of the independent variable and no real control/reference condition. One way to think of the Zimbardo study is that the research question is: "What will subjects do as guards under conditions (a) where they receive little or no supervision for compliance with rules for treating prisoners (this is what Zimbardo did) and (b) where guards receive supervision for behaving according to the guard role (after all, in real life prison guard do have to follow some rules). I believe the guards would behave differently if they received feedback on how humanely they treated the prisoners but perhaps this might have detracted from the shock value of Zimbardo's study. -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=16972 or send a blank email to leave-16972-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
