Research Design how to get the data to help you test your hypothesis and draw correct inferences Good research design has 3 properties 1 Avoids threats to valid inference prevents issues that might stop you from making valid inferences Identifies an appropriate unit of observation Identifies special and temporal domain 2 3 1 Random selection chosen give the treatment to randomly 2 Random assignment the units that end up in your discriminant are randomly once you ve chosen your sample you choose who to 3 ways to maximize comparability 1 Standard experiment 2 Statistical control having an effect 3 Matching use two groups that are different only in 1 aspect you use stats to control for other possible things that are never technique tries to choose groups that are exactly the same In addition to standard experiments there are Field experiments you actually go out into the field and conduct the experiment Natural experiments experiments that get lucky Yes this is what he said One of the benefits of doing research design is to avoid for us to evaluate Two types of validity 1 Internal validity the change in the dependent variable outside of the research itself 2 External validity does the change in our independent variable really cause how much your experiment research has relevance Experiments also help you control for stuff that you observe but don t take into account Convenient samples who you can get for your experiment A proper experiment eliminates hypothesis Most political science is not experimental instead we use observational studies Observational Study where you observe the data that are out there as they are You have to explicitly include everything that you think is important in an observational study Theories are conceptual hypothesis are operational Conceptual the theory itself is conceptual definitions of the theory are conceptual Operational something capable of being measured In order to maximize conceptual clarity we want to 1 Define characteristics and boundaries of the concept 2 Know the unit of interest 3 Know the variation Error is inevitable you will never be able to produce an error free study Three error related concepts 1 Reliability 2 Bias 3 Validity Systematic error always too high or too low test retest alternative form asking the same question in a different way split halves Face validity is it valid on the face of it Content validity does your measure have all of the necessary elements Construct validity does my constructive measure match other constructive measures of the same concept internal validity are you measuring your concept appropriately Discrete measure can count the number of things that there are Continuous measure take any number you want Income is an example of a continuous measure This is all he said about these two last class Types of data Categorical Nominal Bunch of categories can t order them race gender etc Ordinal Categorical but you can rank them highest education received Interval ratio numbers you can order them and they have actual values Nominal and ordinal are categorical Population The set of individuals items or data from which a statistical sample is taken A sample is a subset of a population The bigger the sample the lower the error A lot of the error we get is from sample error Describe your data used with descriptive statistics Descriptive statistics is a central tendency Measures of a central tendency are mean median and mode Mean relevant measure you care about UNLESS there are outliers
View Full Document
Unlocking...