POS 3713 Midterm Exam II Study Guide 1 What is a research design and what are its chief goals a Research Design The strategies that a researcher employs to make comparisons with the goal of evaluating casual claims b Types of research i Exploratory ID new facts ii Descriptive Describe observables in world iii Explanatory Theoretically explain causal relations phenomena of interest c What it is i Formally 1 Specifies the procedures we will use to answer a research question 2 What data will we use Time Space 3 What hypothesis tests most appropriate ii Informally 1 Ensures that data we examine provide an answer to research question d All research designs specify comparisons i Comparison Fundamental building block for causal inference 1 May compare values of y across x s units 2 May compare values of y over time e Wants groups to be identical on all variables except for the IV of interest i Maximize comparability 1 Randomization Randomly assign units to values of x IV 2 Statistical Control If we suspect other variables z are also moving y we can include them as additional IVs holding them constant 3 Matching Select cases into dataset similar on all variables except IV of interest f Goals i Allow for valid inferences as to causal effects of x on y ii Good research design will account for 1 Threats to causality 2 Unit of interest 3 Temporal spatial variation of interest iii Goal Compare like with like 1 To assess whether we may infer that x IV moves or changes y DV we must compare two or more units 2 Ex We suspect that rainfall IV on Election Day influences voter turnout DV by precinct We must compare rainfall turnout in one precinct to another precinct where rainfall is different Did precinct with less rain have higher turnout iv Goal Answer the how else question 1 In comparing dependent variable values between two units individuals etc how else might those units differ from one another 2 Ex Comparing two precincts possible that one has more wealthy residents that are less discouraged by rain wealthy residents do not take public transportation in bad weather 3 To infer a causal relationship want the units of interest to be as similar to one another as possible except on the independent variable of interest 2 What is an experiment How does it differ from an observational study a Experiments Research designs in which the researcher both controls and randomly assigns values of the independent variable to the participants i An experiment controls and randomly assigns x 1 Control Value of the IV determined only by researcher ii Experimental Design 1 Researcher randomly assigns subjects to one of two groups a The experimental group b The control group 2 Researcher controls the introduction of the experimental factor 3 Measure the two groups on the dependent variable of interest 4 Compare the two groups If there is a difference you can attribute it to the experimental factor b Observational Studies Research designs in which the researcher does not have control over values of the independent variable which occur naturally it is necessary that there be some degree of variability in the independent variable across cases as well as variation in the dependent variable i Take the world as it is and studies naturally occurring differences between units ii Researcher does not have control over values of the IV iii Must have variability on IV and DV 3 What are the strengths and weaknesses of experiments Observational studies a Experiments i Strengths 1 Are particularly good at crossing the 2nd and 4th hurdles b 4th Is there some confounding variable z that is related to a 2nd Could Y cause X both x and y 2 Have high internal validity a Researchers are confident that the effect of x on y is real ii Weaknesses 1 Not all IVs are controllable 2 Ethical considerations prevent some experiments 3 Low external validity b Observational Studies i Strengths 1 Have high external validity a Researchers are confident in the relationship between x any y beyond the sample ii Weaknesses 1 Internal validity Could x be causing y 2 Ecological Fallacy Drawing inaccurate conclusions about individuals based upon aggregates 3 Regression to the mean Are extreme values of y due to chance 4 What are field experiments and how do they compare to traditional experiments and observational studies a Field Experiments An experimental study that occurs in the natural setting where the subjects normally lead their lives i Researcher tries to influence value of x 1 May not have complete experimental control 2 May not be able to use random assignment ii Behavior of interest observed in natural setting iii Advantages 1 Natural setting high external validity 2 Longer observation periods iv Disadvantages 1 Do not have complete experimental control 2 May not be able to use random assignment a Therefore treatment and control groups are not balanced on Zs b Traditional Experiments i Researcher assigns values of x ii Using random assignment c Observational Studies and Field Experiments aren t fundamentally different 5 What is the difference between random assignment and random selection What types of studies use each How does each contribute to internal or external validity a Random Assignment Values of IV assigned randomly i Participants randomly assigned to treatment or control group s ii Random assignment helps ensure all groups are identical on all z variables iii Experiments b Random Selection Members of study selected randomly from population i Also known as random sampling ii Random sample of people iii Differences between the sample and the whole population 6 What are the two types of observational studies Why would a researcher choose to use each one a Two types i Cross sectional study Many units sampled over a single time period hold constant other factors ii Time series study Single unit sampled over many time periods 7 What are internal and external validity How do they differ What are some common threats to each one a Internal Validity Confidence that the effect of x on y is real i The degree to which a study produces high levels of confidence about whether the independent variable causes the dependent variable b External Validity confidence that the relationship between x and y generalizes beyond the sample i The degree to which we can be confident that the results of our analysis apply not only to participants and circumstances in the study but also to the population more broadly construed c Experiments that have high internal
View Full Document
Unlocking...