Lecture notes, Everyday vs. Scientific Inquiry PDF

Preview

Summary

- merged files: soc 293 notes.docx - soc 293 2 notes.docx...

Description

1/28/16 SSC 293 Notes Everyday vs. Scientific Inquiry 





In everyday situations, we do a poor job analyzing poor quality information and this arrive at a poor understanding of the world o Resolution  Recognize how we come to know things and what we really do not know  We usually think we know things because of tradition or authority  Avoid common mistakes  Inaccurate observation (not paying attention): UGA students – how many columns are there in front of Baldwin Hall  Overgeneralization: My neighbor from NYC is rude; people from NYC are rude. o Get large representative samples, replicate study  Selective observation (“confirmation bias”): meteorologists are horrible at their jobs. o Only attentive to times someone is wrong o Know underlying bias or seek cases that don’t fit underlying assumptions  Illogical/faulty reasoning: “gamblers fallacy”, coin flipping, inferring cause and effect from correlation  TRADITION & AUTHORITY - WE USUALLY THINK WE KNOW THINGS DUE TO THIS “Health Statistics May be Bad for our Health” 1. Fear/”moral panics” 2. Anchoring (round numbers and attachment to initial numbers) 3. Correlation is not causation (faulty logic) 4. Put numbers in context (what do they mean, how were they obtained) a. Rate is determined by “number of teen girls/ 1,000 teen girls” i. It can only change if number of teen girls changes b. Proportion of births in teens “number of teen births/all births” i. This number can change if older women have fewer children What is the goal of social science o Not about establishing scientific laws o Uncover and explain patterns of regularity  Some patterns are common sense but have big implications  Birds of a feather: Homophily

Unlike scientific laws, we can consciously disrupt patterns (and/or their negative effects)  Plenty of room for contradictory cases! Probabilistic vs. Deterministic Arguments o Claim: (On average) The undeducated are more (likely to be) prejudiced than the educated – must read sentences as probabilistic  This claim doesn’t mean that every single educated person is unprejudiced BUT the pattern still holds. 



2/2/16 I.

II.

“Education increases wealth” a. Deterministic or probabilistic? Det. b. Prob- on average education increases wealth It is also helpful to distinguish between two diff kinds of explanations c. Idiographic explanation – an exhaustive explanation of a specific case (think idiosyncratic) i. E.g. why did you come to UB vs somewhere else? (specific reason) d. Usually linked to qualitative research e. Often useful for illustrating a point or generating ideas f. Nomothetic explanation – a parsimonious explanation of a class of events (always a partial explanation; always exceptions) i. E.g. – why do students go to UB g. Usually linked to quantitative research h. Often used for testing a general rule and explaining patterns ( but always incomplete) i. R^2 : 1= best but we get excited when we get a .3 j. Use both but use them carefully and for diff. goals.

PARADIGMS, THEORIES & CAUSAL DIAGRAMS I. Paradigm – fundamental models or frames of reference used to reason a. General frameworks b. Common in assumptions about the nature of reality c. Common sociological paradigms: conflicts (including critical race and feminist, functionalist, symbolic interactionists d. E.g. feminist: gender is a social structure, a primary organizing principle of identity, interaction and inequality. II. How does a theory differ from a paradigm?

WHAT IS A THEORY? I. Theory: an explanation of how and why at least two concepts are related. A theory is more than a descriptive story or a general orientation because it explains something II. “ wealthier people tend to be healthier than less wealthy people.” – Theory ? NO there is no “because” A THEORY I. in words: social class is related to better health because it is associated with greater access to higher quality foods Class > (+) access to high-quality foods> (+) Health II. how does a theory differ from a hypothesis? a. We must operationalize the concepts in a theory in order to make a testable hypothesis.

OPERATIONALIZING CONCEPTS Theory : Class > (+) access to high-quality foods> (+) Health Testable Hypothesis: people who earn a higher weekly salary(class) will eat more fresh fruits and vegetables(access) and therefore will have a lower BMI(health) Weekly salary 

>

+Servings of fresh fruits * veggies >

-BMI

2 diff between theory and hypo – we use boxes in hypo & operational concepts

KEY POINTS for CAUSAL DISGRAMS 1. THEORIES = CONCEPTS = CIRCLES 2. Hypothesis = operationalized concepts = boxes 3. Every causal model must have: arrows, signs, boxes/circles 4. Let your signs tell u ur directions ( multiply across a chain 5. Interpret each pair of boxes first and then the whole diagram. BOTH MUST MAKE SENSE 6. CATEGORICAL VARIABLES (E.E.G GENDER, RACE) MUST BE SPECIFIED SO that the model is interpretable.

2/4/16 Deducting using data  Topic: why do some kindergarteners have trouble spelling their names?  Theory: longer names lead to difficulty spelling because they are harder to sound out

o Step 1 : Draw a causal diagram of a testable hypothesis that is derived from the theory above #of letters in name ----- spelled name right on first day  Step 2: evaluate my testable hypothesis using data Inductive reasoning 

Up the pyramid ( data  Hypothesis  Theories  paradigms

Pros and Cons  Deductive reasoning o Advantages – Data help test theory o Dis- limits scope of questions  Quantitative work – (surveys, experiments) – usually nonmytheical explanations. Patterns. Deductive o  Inductive reasoning o Adv- generate ideas o Dis- theory not tested. The theory always fit the data. o Qualitative- in depth limited explanations. Help generate ideas and build theory.

2/9 

 

Principles of experimentation o Voluntary participation  No pressure, bribe, withdrawal at any time o Do no harm  Stress, physical o Confidentiality/anonymity (not even researcher can identify you) Books do not undergo peer review process ABC and Burger. . tried to replicate o 150 volts max o All participants were evaluated o Participants were told 3x ( cant leave at any time ) before and during o Debriefed immediately o Who did the study was a psychologist and was told to look for any symptoms of stress and then immediately end the experiment.

CAUSATION  Three criteria for causation o The cause and the effect must be correlated o The cause must precede the effect o The relationship cannot be spurious



Are the cause and effect correlated o If they are, the value of one is related to the value of the other o A positive correlation- same directions o Negative – opposite directions

 

Correlations can be strong or weak DOES THE CAUSE PRECEDE THE EFFECT o Causal order is not always clear from cross- sectional data. Use theory and logic to evaluate when possible o X= independent variable=cause o Y= dependent variable= effect COULD THE RELATIONSHIP BE SPURIOUS? o THE association between X and Y cannot be a mere correlation due to the fact that the cause ad effect are both caused by a 3rd variable. ESTABLISHING CAUSATION USING DATA o Proposition  Hard work leads to financial success. In other words, financial success is caused by hard work. o Operationalization  Use deductible reasoning to make a testable hypothesis  Hard work = currently has a FT job  Fictional data source: random sample of US adults collected at one point in time For something to be spurious – 3rd factor must cause both things.





 





ADDITIONAL CONSIDERATION o Are the operational definitions in our hypothesis valid measures of hour theoretical concepts o Are we talking about a sufficient cause or necessary cause? VALIDITY o Do the measures(operational definitions) we used allow us to conclude that the concepts are related? That, are our measures valid in that we are measuring what we think we’re measuring? o Concepts – Hard work + financial success o Measures- FT JOB + $30,000 +in the bank o To the extent tht our operational definitions are not valid measures of our concepts, our ability to support the theory will be limited. NECESSARY VS. SUFFICIENT CAUSES o Necessary cause: condition must be present for the effect to follow o “if no X then no y” o E.g. if you are not female, you will not become preg.  Are there other avenues to being preg ? no  Will all woman be preg. o Being female is a necessary cause but not a sufficient cause

 CONCLUSIONS o Establishing causation is often difficult because  Time ordering can be hard to establish  Associations are often imperfect  Spurious relationship mislead us  Our measures are often important

2/16 CONCEPTUALIZATION & MEASUREMENT   

Concept. . – this is the process of defining what a concept means. For instance, what does it mean to be healthy? We can define indicators(BMI, BP) of our concept and dimensions (e.g. mental, physical, social health) of our concept. Application: think about what it means to be “intelligent” and then write a normal definition ( like a dictionary definition) o Being able to apply things you learned to real life in hopes of bettering oneself.

MEASUREMENT This is the process of figuring out how to recognize a concept in the real world. This process is also called operationalization.

ISSUE 1 : VALIDITY (ACCURACY) Are you measuring what you think you are measuring? Different ways to assess validity 1. Face: do you think measure is good? Does it seem to make sense? 2. Content: have you measured all the important dimensions? 3. Construct: is it related to other theoretical concepts in the way it should be? 4. Criterion: does it predict some outcome that is expected to predict well? ISSUE 2: RELIABILITY (PRECISION) When you measure something, are you likely to get similar results over and over? Threats to reliability 1. Cant answer: how old were you when you spoke your first complete sentence? 2. Consider irrelevant: what do you think about fishing restrictions? 3. Complicated: how many phone numbers do you have memorized? Safeguard

 Test retest  Split half  Use of established measures TENSION Reliability and validity are foten related in a negative fashion. The more accurately you measure something, the more chance there is for error. This and rest are ublearns.

REVIEW Nec. Cause : If no y then x Suff. Cause: if x is there y will follow Idiographic – IRB must review 1. Voluntary participation 2. Do not harm 3. Confidentiality

2/25/16 Measurement challenge  Single indicators o Does not usually tap all the dimensions of a concept. o Do not usually measure intensity very well SOLUTION 1: QUESTION FORMAT  Likert question format : (strongly agree, agree, disagree. . . )  Semantic differential format (identify agreement with too opposite dimensions [ boring to exciting 1-10]) SOLUTION 2 : MULTIPLE INDICATORS  Multiple indicators o Can measure multiple dimensions o Provide a more detailed indication of intensity o Recall: Poor-self control items Concept:religionsity  How many times do you attend religious service (e.g.of multiple indicators)  How religious are you? (1-5)  How important do you consider religion to be your everyday life? INDEXES AS A SOLUTION  Example 1 : an idex for a “good occupation” (see worksheet  2xample 2 : . . INDEX CONSTRUCTION  4 STEPS o select items o examine their empirical relationships o Score index o Validate the index 1. SELECTING ITEMS  FACE VALIDITY o Is there general agreement that the items tap the intended concept? o Items should be logically related to one another. 2. EXAMINE EMPIRICAL RELATIONSHIPS  Evaluate bivariate relationships while constructing an index. Each item should be correlated with one another. 3. SCORE THE INDEX  We can score an index in several ways  We can construct counts, averages, percents, etc.



How would you make an index with the self-esteem questions in your handout?

4.VALIDATE THE INDEX  After making an index, evaluate it using construct validation: o Is the index related to other items(indexs) that measures popularity? SCALES AS A SOLUTION  SCALES o Allow individual items to be weighted diff. o Can do better job of measuring intensity and preserving info.

1. 2. 3. 4.

But how should you determine the weights Make them up yourself Bogardus distance scale Thurtsone scaling Likert scaling

BOGARDUS SOCIAL DISTANCE SCALE EXAMPLES o Are you willing to permit sex offenders to live in your country? o Are you willing to permit sex offenders to live in your neighborhood? o Are you willing to permit sex offenders in your city ? o Are you willing to permit sex offenders to live next door? o Would your let your child marry a sex offender? LIKERT SCALING WEIGHTS DO NOT CONFUSE THIS WITH LIKERT RESPONSE  USES INDEXES VALUES TO WEIGH ITEMS  EXAMPLE ON WORKSHEET  Advantages – cheaper & faster & relying on your own respondants ( not histocially significant)  Disadvantage – akward to calculate.

3/3 SAMPLING CATCH UP  What are the two main types of sampling? o Probability vs nonprobability  What process distinguishes the two?



o You randomly selected our respondants.if we going to collect a probability sample, we need to a sample frame ( list ) , randomly select ppl. In non prob, we don’t do any that. What distinguished them in their capabilities o Probabilities can generalize.

LEGGINGS AND BOOTS AT UB PROBABILITY SAMPLING  Different mechanisms of probability sampling  Equal probability of selection (EPSEM) o Smple random sampling (random lists of #s, do down list) o Systemic sampling( every 5th, element) EVERYBODY HAS AN EQUAL CHANCES OF BEING INCLUDED.  Stratified sampling (EPSEM from within strata, like college class, race, gender, etc.)  Multistage cluster sampling ( randomly sample larger elements and then smaller elements within them)  Weighting (oversample certain subgroups and then statistically correcting for oversampling)

3/8 WHEN ARE EXPERIMENTS USEFUL?  Experiments involve o Taking “action” ( or manipulating a variable) o Observing consequences of that action/manipulation  uSEFUL WHEN:  we want to isolate cause and effect  our propositions are not complex/contain few concepts. What 3 keys conditions are necess. For cause & effect? 1. Cause b4 effect 2. Must be related 3. Cannot be spurious Classical design  Fulfill all 3 criteria ( correlataion, time order, & non-spurious) Classical experimental design 1. Independent(cause) and dependent (effect) a. Independent variable typically taes the form of a stimulus (either present or absent) 2. Experimental control group

a. Exper group = group that receives stimulus b. Control – group that does not receive stimulus 3. Pretest & posttest 4. Measure dependent variable before stimulus is introduced and again after for both groups How do we ensure that both control and experiment group start the same?  Randomization & matching VALIDITY: INTERNAL AND EXTERNAL  Internal validity ( is our effect actually due to our cause?) o Threat to internal validity Is present whenever anything other than the experimental stimulus can affect the DV o Often due to improper experimental design  External validity ( can we generalize to the real world)  Threat to external validity is present if results might not be generalizable to ‘real world” o Uniqueness of sample ( cause/effect established in college students might be diff. in general pop) o Causes are rarely isolated in the real world  Experiments are often LOW on external validity ( think about sampling) Threats to internal validity  History  Maturation  Testing  Selection bias  Experimentation mortality  Compensation  Demoralization  Hawthorne effect Layout of a research article 1. Abstract 2. Intro 3. Literature review 4. Method 5. Results 6. Discussion What do you need to know 1. What is the research questions 2. Why should we care (significance) 3. Whats the method? how do the authors go about answering the questions 4. What do they find?

3/10 internal vs external validity ? every method can and cannot tell us certain things

WHEN ARE SURVEYS USEFUL?  Describing population too large to observe directly  Measuring attitudes/orientations of large pop. (think nomothetic)  Requires probability sampling to do well. Good surveys are ver difficult to construct and implement Population Sampling frame Sample Respondents (to questionnaire) Respondents ( to items) we can actually test rates QUESTIONS ABOUT QUESTIONS CONSIDERATIONS  OPEN VS CLOSED ENDED QUESTIONG ( YOU HAVE ANSWERS AND THEY CHOOSE ONE)  THIS IS WHAT WE WANT BECAUSW THIS ENABLES TO ANALYZE OUR DATA.  Closed ended questions should be o One prob about close , ppl will have to choose an answer and the answers given may not be their answer. o Exhaustive - ^, all of the possible answers are provided o Mutually exclusive – respondents should not be able to plac themselves in more than one answer. No overlap.  Items should be clear ( no ambiguity) o E.g. how do you feel about the way America is headed o How good is your child?  Avoid double-barreled questions and answers o E.g. how satisfied are you with your marriage and your job?  Avoid biased items and terms o Social desirability o “don’t you agree’. . . o words with very strong, negative/positive, connotations are problematic (e.g rape, abuse, welfare)

MEASURING RAPE 1. have you ever attempted unsuccessfully to have intercourse with an adult by force or threat of force? 2. Have you ever had intercourse with someone by force or threat of force? HOW TO ADMINISTER?  Self-administrations o Internet surveys  Face0to-face survey o CAPI ( computer assisted  Telephone survey o CATI Surveys weaknesses  Hard to contruct  Limited depth= standardization o High on reliability but low validity  Inflexible  Response bias  Social desirability bias  Artificaity ; are attitudes connected to behavior? In general survey research is generally strong on reliability but low on validity/

3/22 SURVEYS: GENERAL WEAKNESSES  Very difficult to construct  Standardization= limited depth o Remember, survey res. Is high on reliability but low on validity. o Are attitude connected to behavior o Are we measuring actual/beliefs or “meta” attitudes or beliefs?  Relatively inflexible o Particularly problematic with secondary data analysis  Response bias  Social desirability bias SOLUTIONS TO POOR QUESTIONS  Pretest questionnaires o Conduct experiments & focus groups with regRDS TO EFFECT OF QUESTION WORDING o IMPORTANT BECAUSE CONSISTENCY REQUIRED FOR COMPARISON (E.G. LONGITDINAL STUDIES)



Acknowledge and adjust bias in our survey names, questions, and response options o Exhaustive and mutually exclusive options o Clear questions (non-ambiguous. Ne negatives, no double-barreled questions)

PRACTICE WITH EMPIRICAL RESEARCJ (college men rape article) 1. What is the research question 2. Why care? Why is the answer important 3. Method 4. What did they find? a. Can we differentiate? Yes b. Most men endorses neither force nor rape CRITQUE  Method appropriate for research question?  Sampling approach – limitations?  Conceptualization, operationalization,validity, and reliability issues?  Social desirability? Response bias?  Is it a causal argument ( x causes y) and if so do they have necessary evidence. FINAL NOTES  Goal of survey research: simplify data in order to make genralizations about populations  Must strike balance between own worldview and respondants absolute freedom to write whatever they want  Take protective measures to ensure balance is met o Be reflexive about questions/answers o Pre-test surveys/ hold focus groups o Exmine existing literature o Etc. QUANTITATIVE V QUALITATIVE Qualitative  In ...