Example: Why do women hold infants with the left arm?Classifiying research methods Empiricism-objectivity-reliabilityHow correlation is measuredRANDOMIZED DOUBLE BLIND PLACEBO CONTROLLED CLINICAL TRIALPsy 100 Spring 2003 MadiganPsychology 100 Spring 2003 Research Methods Lecture 01-16-03 1. Psychology as an Empirical Discipline Example: Why do women hold infants with the left arm? Classifiying research methods Empiricism-objectivity-reliability Going beyond what's given Challenges to psychological research 2 Experimentation Randomization & Random sampling Control The randomized, double-blind, placebo-controlled clinical trial 2. Correlation, causation, pseudo-experiments How correlation is measured Causal inference from correlation4. Research,Society, Ethics5. What you should knowThe Honest, Empirically-Based Expert- I have studied alternative hypotheses about the nature, causes, and possible amelioration of the problem.- I have evaluated these hypotheses in light of the existing empirical evidence.- I have tentatively concluded that some of these alternative hypotheses are better supported than are others.- Therefore, I understand something about the problem and how to address it, although new findings may always prove me wrong.Be (properly) critical of psychological research!- Is this a fact or an inference?- What alternative explanations are there?- Are causal conclusions based on true experimental evidence?- Have the results been independently replicated?- What might limit the generality of the findings?- Is this a primary or secondary source? What you should know- Be able to classify a given piece of research: is it an experiment, a correlational study, etc. - Know what the correlation coefficient means.- Understand the problems of causal inference from correlations.- Be able to identify independent and dependent variables.- Understand the nature of and need for randomization and random sampling, and the difference between them.- Be able to see the need for different kinds of controls in experimentation.EXPERIMENTATION AND CAUSATIONVariable X is a cause of variable Y if Y varies when X varies and all other relevant conditions (other potential Xs) are held constant. This is the logical principle that underlies experimentation. THE RANDOMIZATION PRINCIPLEWhen you compare an experimental group and a control group you want to be sure that the two groups don't differ systematically in any way except for the independent variable. Problem: people differ in innumerable ways that may be relevant to the dependent variable. They can't be "kept constant." Solution: random assignment. LetPsy 100 Spring 2003 Madiganthe outcome of a chance event (e.g., coin-toss) dictate who gets what treatment. This way, no characteristic of thepeople being tested systematically influences the outcome. Randomization eliminates potential biases.RANDOM SAMPLINGRefers to how subjects are selected in the first place. Purpose: to get a sample that is representative of the populations youwant to generalize to; provides have a rational basis for making an inference from a sample to a population.RANDOMIZED DOUBLE BLIND PLACEBO CONTROLLED CLINICAL TRIALThis is the “gold standard” – the best method for determining the true effects of a treatment or procedure or program. Now widely used in medical research and drug studies. Example: the effects of Paroxetine on socially avoidant behavior.R-DB-PC-CT is an attempt to closely follow the fundamental principle of experimentation: vary one thing, keep everything else constant. Compare this to the “case-control” method: just compare people who use the drug withpeople who don’t. This method usually involves many sources of bias.Example: Evaluating the effects of electro-convulsive therapy (ECT) on depressionSuppose a depressed patient is administered an ECS treatment and seems much improved later. How do you knowit was the shock-convulsion that was responsible for the change? There are alternative explanations: spontaneous remission; effect of drugs administered prior to ECS; patient's expectations; doctor's expectations. You need to perform a controlled experimental evaluation. Assign depressed patients at random to one of two conditions:Experimental Group Control Grouppatient sedated patient sedatedpatient unconscious patient unconsciouselectrodes attached electrodes attachedcurrent delivered no current delivered (or sub-critical current)Also, have the patients evaluated by someone who is "blind" to the treatment received.This will answer the question: is the shock-convulsion a necessary component of the treatment?Psy 100 Spring 2003 MadiganCAUSATION AND CORRELATIONProblem: it is often difficult or impossible to perform true experiments in psychology: a large number of variablesto control; practical, legal, ethical constraints on what can be done (such as randomization). But people still want.to make causal inferences from pseudo-experiments and correlational research. Very important to understand thelimitations of such research. If variables X and Y are correlated, then: (1) X is a cause of Y; or (2) Y is a cause ofX; or (3) X and Y are the effects of some third variable Z; or (4) Y is directly caused by Z and only indirectly
View Full Document