Lecture 5 Notes
Lecture 5 Notes 76884
Popular in PSYC4201W
verified elite notetaker
Popular in Psychlogy
This 7 page Class Notes was uploaded by Rachel Onefater on Tuesday January 26, 2016. The Class Notes belongs to 76884 at George Washington University taught by Dr. George Howe in Spring 2016. Since its upload, it has received 65 views. For similar materials see PSYC4201W in Psychlogy at George Washington University.
Reviews for Lecture 5 Notes
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 01/26/16
Causal Reasoning We can infer that A influences B: • When changes in A are followed by changes in B, • given that absence of or differences in the pattern of change in A would have been followed by absence of change in B, • all other plausible causes being equal. → In the chapter, Howe talks about “all other things being equal” is quite IMPOSSIBLE. Ex: giving you a medication to see how your heart responds and at exactly the same time, he does not give you medication to see how your heart responds→ Does not happen in reality → We can infer that something causes or influences something else. What do we mean by “being equal”? • Operating in the same way in both the factual and the counterfactual condition • In experiments: – There can be other things influencing B, but theycan’t diffe for the experimental and control groups 4 Example: Williams et al • Researcher assigned people to one of two conditions – Verbal worry → Worry, but use words – Imagery worry → Worry, but have images and pictures in your head • Then measured attention bias towards threat words (dot probe task) • Could other factors (personality, recent experiences) influence responses in dot probe task? Answer : personality, recent experiences, and when they come into the study, it will influence whether or not they were attentive to something. → Ex: saw car accident, ended a relationship, comes from an abusive household ro a volatile household, someone has some sort of mental condition i.e. Asperger's Syndrome, inherent sensitivity to stimuli Selection? • What if people had been allowed to choose their condition (verbal or imagery)? → Would suggest that there are things we experience/something about who we are that could influence that choice! • What factors (personality, recent experiences) might influence their choice? Answer : For young kids, boys tend to develop more slowly verbally than girls do, vision problems might make imagery harder for that individual. Selection effects • Could any of the factors we just came up with also influence sensitivity to threat? Answer : Sensitivity to contact, some people may be more sensitive to threat cues, verbal threat cues may be more pushed away from the verbal condition, so you get more people from the imagery from the imagery condition, and they become more sensitive to threat cues. → Personality style pushes someone to select a certain style(imagery/verbal), and then affects how they respond to threat cues. Selection effects • If yes, our groups are not meet the “all other plausible causes being equal” requirement → People in them(groups), share the same set of causes! A visual tool • Directed graphs • Labels are causes or effects • Arrows indicate direction of cause Mode of worry → Sensitivity to threat Various possibilities – (Possibility 1) The way I think it goes Mode of worry → Sensitivity to threat – (Possibility 2) The way I think it goes, but with selection factors influencing both *NOTE: Selection factor(influences both of the factors below) Mode of worry→ Sensitivity to threat – (Possibility 3) I’m wrong, and only selection factors are causal Selection factor → Mode of worry Selection factor → Sensitivity to threat If groups differ on selection factors, which of these possibilities are consistent with : • Findings that those in the verbal condition show faster threat sensitivity than those in the imagery condition? Confounding • Confounding occurs when some third factor: – Is associated with my supposed cause – Influences the supposed effect • When confounding occurs, any observed association between supposed cause and effect could be purious(def.) wrong → Faster in the verbal condition, and slower in the imagery condition, and by allowing people to select in, experimenter has created a confounding study. Any association between the two variables could be incorrect, but there could be no way of knowing if you are wrong/right! What if a person gets both conditions? • Personality, recent conditions are all the same, and they can’t be confounds – “A person acting as their own control” • BUT: The problem of spillover – If I first use verbal worry, could it change the effects of later imagery worry? *NOTE: When you are in the second condition, your history is no longer the same because you were in the first condition first! (iwithinsubject study). • Example: Trials comparing two medications – If someone gets both meds, need washout period to be sure first med is out of the system, and then have them come back after a time when you know it has washed out of the system → need to have prior theory, understanding, and studies, and may mean no longer detectable levels in the blood stream, or other ways). → Anesthetics have an effect up to a few months later, so this can be a complicated method! Can one person be the control for another? • Juan does verbal worry(happy person), Jenny does imagery worry(angry person) • Personality, recent conditions are very likely to be different Answer : No! Not useful at all Can one group be the control for another? • Personality, recent conditions are more likely to be equivalent on average → When you get larger groups, the average attitude, shifts toward the middle, and average attitudes are going to be similar under certain conditions! Experimental counterfactual: Average causal effect (picture on slide) Intervention to Learn Stress management techniques: Sleep quality (No treatment vs. Treatment) Counterfactual assumption: Group averages equivalent on all other plausible causes → Sensitivity to context, but you look at the measure of sensitivity context in group, they should have pretty much equivalent average, group difference cannot account for the treatment effects as long as you have an equivalence! What research conditions support this assumption? • Various research methods available for getting closer to equivalence on confounds • Depends on type of confounding Types of confounding • Selection effects – Some third factor influences selection into our supposed causal conditions, and also influences our outcome → Have been a very important for trials of intervention/programs (Example: very first study of Head Start to see if it had an impact on kids development of educational skill). • Chance effects – People in our causal conditions differ by chance on some factor that influences the outcome • Induced confounding – We think our conditions vary only on the feature of interest, but they also vary on some other feature that influences the outcome. → You create the confounding as you go along, and Prof ill ask us how to design our studies not to do that! How to eliminate selection effects? • Use a selection method that can’t be biased (by either your participants or you) → MAKE it completely chance! – Random assignment (each person has equal probability of being in one group or the other) – When random assignment is used, groups will be closer to equivalent even on unknown or unobserved factors Can groups differ on confounds by chance? • Yes, even when using random assignment – More likely with smaller samples – Ways to improve on simple randomization • Therefore, good to measure potential confounds and test whether groups are equivalent or not – Can adjust statistically if differences are found What if you can’t use random assignment? • We often aren’t able to randomly assign people to causal conditions 1. Ex : Teen pregnancy on the development of a child at age 4→ cannot assign people to be pregnant or not! 2. Ex: EMT member are sent to a brutal tragic accident, and you can see if this affects later PTSD 3. Ex: Test people who smoke on Cancer development 4. Ex: GENDER is a big one too! If gender is a viable, alternate plausible reason for people to respond to your measurement. • Here, identifying, measuring, and testing potential confounds becomes the main method • But, how do we know what factors to look at? Plausibility • Don Campbell : major methodologist in social scientist suggested that while there are many possible confounds, what we are really interested in, is in those that are plausible • Focus on plausible alternative causes of B that could also vary by A (i.e. confounds) What is Plausible? • How do we determine what is plausible? – Common sense – Prior research – Theory • Example: trait anxiety has been found to be associated with attention bias to threat – So groups in Williams study need to be equivalent on average levels of trait anxiety, so this was based on both theory and on prior research • What is plausible can change as science in a specific area progressescan never prove anything with 100% certainty! Induced confounds • In experiments, we create both factual(experimental) and counterfactual (control) conditionsdesign materials that you expose people to. •Withinsubject example: using dot probe task to test whether threat words lead to faster response than nonthreat words – Suppose we show someone 24 threat words and 24 nonthreat words – They respond with either the left or right hand, depending on the cue (one star or two) → It is likely that a person is faster with one finger than the other finger(more often than not, this is true). – Words appear at the top and bottom of the screen What if the threat words always appeared at the top? What if the two star cue (right hand) was always associated with the threat word? → You have now created a condition where location of the cue is mixed up with whether it is threatening or not! Induced confounding • Creating conditions that differ on factors other than the ones we are interested in → Words can differ in threat meaning, length, and how difficult they are Ex: Hate and Supercalifragilisticexpialidocious, should not be compared to one another • When those factors have some impact on the outcome of interest How to reduce or eliminate induced confounding? • Creating conditions that differ on factors other than the ones we are interested in • When those factors have some impact on the outcome of interest How to reduce or eliminate induced confounding? • Design experimental and control conditions so they differ on your construct of interest, but not on other factors that could plausibly influence the outcome • Lab 2: you will be design both experimental and control conditions for a study: – How your conditions vary – How your conditions need to be similar Example • Experimental trial of yoga for stress reduction – You believe the active ingredient is explicit training on how to gain mindful awareness of muscle movement, position, and sensation – Design a control group that is similar in terms of: • Number and length of sessions • Sequence of yoga poses in each session • Level of stretching and strenuousness – But differs in having no explicit mindfulness training , want to rule out as potential alternate effects that you do not want to induce. • Notice that this tests the influence of the mindfulness training, not yoga in general How do we keep plausible causes equal? • Experimental studies: – Random assignment at various levels – Constrain the stimuli or contexts we manipulate → Cortisol a stress hormone, but your body produces it all the time, and produces on a daily fluctuation and as you wake up, system up, your cortisol pops up pretty fast! Depends on the time of day you do somethinglevel of cortisol changes at different points during the day for different people. • Any study: – Observe plausible alternative causes of B – Determine whether they differ for A – Adjust statistically if they do
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'