TEST 1 Epistemology study of how do we know what we know Traditionally we know things because of authorities tell us intuition precedent common sense and observation Inference descriptive inference ex I want to figure out what percent of pop supports president So I ask percent unemployed In this group half approved half did So that sample would rep the entire population Causal inference seeing relationships in the sample noticing the people who are unemployed are happy with pres and those that are employed are happy with Its basically looking at one subset and saying that s what is going on in the entire world infer a relationship between 2 concepts This requires quantitative and qualitative judgements Ex I believe x caused y I notice x occurs right before y occurs Causal or coincidence What if z causes x and y SO WHAT WE DO IS we make inferences by We cant observe causation We don t see x make y happen We observe CORRELATION We see x and y move together We come up with a causal story to explain why the correlation takes place Prediction using info to predict future value of a concept or variable Correlation 2 var seem to move together Causation changes in one variable leads to changes in another variable THIS WOULD REQUIRE CORRELATION NO CAUSATION WITHOUT CORRELATION Deterministic relationships X is the cause always leads to effect Y When x occurs y will also occur with certainty Like a formula almost Probalistic relationships X the cause USUALLY leads to effect Y When x occurs y will tend to occur BUT NOT with certainty Stochastic means random Acknowledges there is a random event that takes place that will cause our theory to be wrong every once in a while Science is accumulation of knowledge which is why we care about previous research Or we say previous info was important because it was wrong but not we know what s right you can take previous measures that are valid and proven and build on them Not laws in any deterministic sense Duvergers law majority voting election system naturally leads to a 2 party system Hotellings law in many markets it is rational for producers to make their products as similar as possible Goal of scientific research is inference We cant observe the world and know things because we tend to Inaccurate observations Tendency to over generalize Selective observation Illogical reasoning Fallacy of Affirming the Consequent Just because we have evidence and reason in our favor we havent proved out theory yet ALL WE CAN SAY IS WE HAVE YET TO FALSIFY OUR THEORY If A is true then B is true We observe B and conclude A is true This doesn t not hold logically Ex If in Tallahassee A then your in florida B but if in Florida B then you might not be in Tallahassee A so then you do testing and find out their not correlated therefore not causation If a then b if not b then not a Data is occurences and phenomena Job of pol sci is to systematize date into concepts theorize a story that gives data meaning tell cause and effect Analyze concepts in real world and relationships mustb e testable Independent variable is the suspected cause Dependent variable is caused or depends on influence of outside factor Also called outcome variable Dependent variable is denoted by y Indep is cause Dep is effect Indep Causes dep Var and it comes before or is independent of the outcome they influence Parsimony means simple Truth proven etc Agreed upon terms because they have common characteristics I buy boat you buy boat I buy bigger you buy bigger there is no independent variable Induction to deduction empirics in the world observed then put to theory Deduction to induction theory then observed in real world empirics Causal theory offers a causal story of how and why a concept moves or influences another Should offer additional implications x causes y what else might it influence and how Evidence in support of our hypothesis is simply a failed falsification Hypothesis relationship we expect to observe Deduction Testable statement derived from theory that indicated cause and effect from 2 concepts Wants hypothesis to make strong explicit propositions Should be linked to other propositions specifically independent and dependent var Must say direction of relationship pos neg and type of relationship curvilinear Linear Testable implications of our theories Null hypothesis expected relationship we would observe if our theory is wrong Implicit in al hypoth When we test our hypothesis what we are doing is ATTEMPTING TO REJECT THE NULL HYPOTHESIS Falsifiability ability of hypothesis to fail to be falsified helps to discern the worthiness of a theory If it cant be falsified its not a hypothesis When we falsify the null we only keep our theory safe for now Puzzle a puzzle means I observe 2 things I expect them to have same outcome yet they have diff outcomes Or visa versa Starting with puzzle is good reason 1 it grounds ur theory in concrete cases Reason 2 it requires some knowledge 3rd reason it leaves u to look for non obvious explanations Bivariate one indep And one dep Variable Multi variate has multiple indep Evaluating theories Strengthening All of this below U want to be as Generalized as u can with nouns and verbs but don t over generalize It has to be falsifiable Parsimony simple fewer moving parts elements the better The fewer the more cases u can apply it to Simple doesn t mean easy to understand Assumptions generally are true the irrationality of it makes us think of making choices Rational choice 2 things to know 1 know what ur preferences are 2 choices are consistent Operational def what is the actual things elements you are measuring Reliability ur measures are consistent Validity whether or not the thing ur measuring is the concept ur talking about 2 types of data to look at time series data one case that were observing multiple times Watching our grades each test Cross sectional data one period of time many cases endogenous relationship x causes y and y causes x indep caused by dep and dep caused by indep So you have no direction in endogeneity Spurious relationship x Z Y it makes it look like x causes y ex people who use condems have stds So condems cause stds No people who use condems are sleeping around TEST 2 INFO Research design Formally specifies procedures we will use to answer a research question What data will we use Time series Cross sectional What hypothesis test is most appropriate Informally question Ensures that the data we examine will provide an answer to the
View Full Document
Unlocking...