Chapters 1 3 Theory A tentative conjecture about the causes of some phenomenon of interest Hypothesis A theory based statement about a relationship that we expect to observe There is always a corresponding null hypothesis what we would observe if there were no relationship between an independent variable and the dependent variable Hypothesis testing process in which scientists evaluate systematically collected evidence to make a judgment of whether the evidence favors the hypothesis or the null hypothesis If the hypothesis survives rigorous testing you can gain confidence in the hypothesis and therefore gain confidence in the theory which generated it Empirical Tests Tests based on observations of the real world Evaluation of Hypothesis evaluate the hypothesis from what we found Evaluation of Casual theory from the results of the hypothesis tests evaluate the casual theory Scientific Knowledge confirm evidence or revise Variables do not need to be explicitly defined at the theory level Positive Relationship A relationship for which higher values of the independent variable tend to coincide with higher values of the dependent variable and vice versa Upward slope Negative Relationship A relationship for which higher values of the independent variable tend to coincide with lower values of the dependent variable Downward slope Rules of the road 1 Make your theories casual 2 Don t let data alone drive your theories 3 Consider only empirical evidence 4 Avoid normative statements 5 Pursue both generality and parsimony Identifying Interest Variation Theories are designed to explain variation in the dependent variable identifying this variation is a good starting point First identify the spatial and time dimensions over which the dependent variable is measured o Time dimension identifies the point s in time at which we will measure our variable annually quarterly etc o Spatial dimension identifies the spatial units that we want to measure states nations etc o One of these two dimensions will be static and we will only measure one of the two types as our dependent variable The two types of measures for time series variation o Time series measure the spatial dimension is the same for all cases and the dependent variable is measure across multiple points in time Example Presidential approval ratings between 1995 2005 o Cross sectional measure the time dimension is the same for all cases and the dependent variable is measured across multiple spatial units Example military spending in 2005 across various nations Examining previous research and keep these questions in mind Did the previous researcher miss any other causes of the dependent variable Can the theory be applied elsewhere Are there further implications of their findings How would this theory work at different aggregation levels Form you theory maximizes Formal theorists assume all individuals are rational utility Theories vary be each individuals incentives Measured by utility a vague quantity sum of all benefits sum of the costs of that action When we do not have complete information we use expected utility Denoted by the use of an E around each term To evaluate whether or not some X causes some Y we need to cross the four casual hurdles 1 Is there a credible casual mechanism that connects X to Y 2 Can we eliminate the possibility that Y causes X 3 Is there covariance between X and Y 4 Have we controlled for all confounding variables Z that might make the association between X and Y spurious Finding a relationship is NOT the same as finding a casual relationship o Deterministic Relationships If some cause occurs then the effect will occur with certainty o Probabilistic Relationships causation is normally understood as probabilistic so the effect is not certain Selection Effect systematic error due to a non random sample of a population Or a Z variable that leads to a non random sorting of subjects into different categories of the X variable Example parental involvement in private schools Parents tend to be more involved in private schools vs public schools Chapters 4 5 Designing Research two approaches o Experimental study benchmark of scientific research o Observational Study emulates the first Experiments Research design in which the researcher both controls and assigns values of the independent variable X to the subjects Two components to experiments o Random assignment controls for the values of X and assigns those values to subjects randomly o Control Ensures subjects are divided into a treatment group receive the treatment and a control group receive nothing Because X is caused by randomness we can erase the connection between Z and X Experiments and Internal Validity Crossing the four Causal hurdles 1 Is there a credible casual mechanism linking X to Y o No difference between experiments and non experiments 2 Can we rule out the possibility that X causes Y o It is impossible that Y causes X in an experiment because X is assigned before Y is measured and if X is generated by randomness alone nothing else can cause it 3 Is there covaritation between X and Y o No difference between experiments and non experiments 4 Have we controlled for all confounding variables Z that might make the association between X and Y spurious o Experiments are equipped to help us answer this question definitely Essentially this all means experiments contain a strong confidence in the casual inferences drawn from the initial analysis If the research produces high levels of confidence on the conclusions about causality it has high internal validity Vice versa Random Assignment vs Random Sampling Random Assignment Subjects of an experiment are assigned randomly to one of several possible values of X the independent variable Random Sampling The way researchers select subjects for inclusion in a study Each is selected at random and has an equal possibility of being selected Drawbacks to experiments to watch 1 Can we really assign X to subjects Social scientists cannot technically assign X as a variable essentially they are assigned by qualification 2 What about external validity Experiments can suffer from low degrees of external validity Degree to which we can be confident the results of our analysis can be applied to the broader population 2 concerns external validity of the subject itself and of the 3 Are there ethical considerations Ethical Dilemmas can plague our research with another stimulus disadvantage 4 The mistake of emphasis Sometimes
View Full Document
Unlocking...