You are on page 1of 13

Sociologists want to explain the hows, whys, whens of social life. Does x cause y? Under what conditions?

Two key designs for evaluating causality Observational Look at different aspects of a phenomenon and see what tends to go with what (if x changes does y change) i.e. tv watching and trusting neighbours

Experimental Manipulate x and see what happens to y. i.e. Make people watch more tv and see what happens to their trust in neighbours.

Experiments while best for causal inference are often impractical and sometimes impossible for social research. Though theyre increasingly used in social policy evaluation. Hence observational designs are far more common. But to judge the adequacy of observational design, you must understand the logic of experimental design.

An experiment Do teacher expectations affect academic achievement? -x = treatment (teacher shown past record of high achieving children) - O = observation of a test score (y) O1 X O2 After teacher made away of spurters , the spurters did better at O2 than at time O1. Did higher expectations cause the better results?

Randomisation & Counterfactuals Problem Its unknown what would happen if the teacher didnt see the past records. Maybe the children with good records would have done better anyway?

Counterfactual We want to observe the scores of a spurter whose teacher knew their record and who didnt know their record (the counterfactual) Impossible as we cant observe this.

So we create two groups by randomisation (random allocation to treatment or control) to simulate the counterfactual of no treament. Comparing the outcome for the treated w/ the control group, we can infer that positive teacher expectations cause children to do better.

Works as the two groups should on average be the same as each other on all observable and unobservable characteristics. Onlu difference at T2 is that one group has received the treatment. (teacher shown spurters past records). T1 Obs1 Obs3 Exp T2 Obs2

No Exp Obs4

Key to know that observational studies (if were trying to explain) should be judged on the same criteria and their limitations acknowledged and mitigated where possible. Longitudinal designs, consideration of all possible confounding influences etc. Statistical control (focused on next year) Types of non-experimental design Cross sectional One shot quantitative survey or qualitative interviews, observations etc

Longitudinal Repeated interviews or observations with same people -Panel design - Cohort design Or different people from same population repeated cross section.

Case Study Detailed estimation of one case (person, institution, family, housing estate). Ethnography.

Comparative Comparison of two or more cases (countries, areas, firms etc) Elements in modern survey research Design of instrument (typically questionnaires) Measurement of beliefs, attitudes, emotions, behaviours, facts

Sample Design Whom do we interview? Distinction of random probability methods and others i.e. quota, convenience, snowball. Cross section/household/panel/cohort

Collection Mode Face2face, telephone, self completion, mail, internet

Analysis

Description, inference, generating and testing theories

Always a practical trade off between costs and quality Total survey error approach *Groves et al+ Maximise quality for a fixed cost. Minimise cost for a fixed quality Survey Error [Groves et al] Answers people give must accurately describe characteristics of respondents Subset of persons participating in the survey must have characteristics similar to those of the larger population Error The extent to which these conditions arent met. Three main types of error Sampling errors (errors of observation) whom you ask Measurement errors (errors of observation) what you ask Processing errors - how you record or process information

Cognitive aspects of Survey Methods approach CASM (Tourangeau et al) Cognitive steps respondents must go through: Comprehend the question as intended What info is sought? Retrieve the required info recall relevant ino Make a judgement/estimate adapt whats recalled to whats required Communicate the answer to generate a recordable response.

Keep things Simple if possible

Complex question more strongly burden the memory - respondents more likely to forget part of the question. So the question ought to be clar. i.e. Now think about the past 12 months, from *DATE+though today. We want rto know how many days youve used prescription tranquilizer that was nor prescrived to you orthat you yook onlu for the experience it caused during the past 12 months Short questions are better on the whole though evidence shows longer questions can help respondents recall past events as above. Use familiar words/concepts Use common words and avoid academic language use the language of the respondent, not yours. i.e. main rather than principal, job-related rather than vocational

Concepts: use in manner in which they are understood rather than in a technical sense.

Avoid/Define ambiguous concepts Common words are often vague so provide a definition, preferably in the question. Do you regularly travel by bus to college during term time? by regularly I mean at least once a week

How worried are you about crime? too vague. During the past week how often have you worried about: Your home being burgled? Being attacked in the street Havin your bank details stolen? Walking alone after dark?

Responses can be Not at all, occasionally, quite often, all the time Leading Questions invite certain answers by giving cues to the respondent How often do you have trouble sleeping? Have you heard of the famous writer John Vooslen? When did you last borrow a DVD? How often do you go abroad? When did you stop beating your husband?

Leading question make respondents feel one answer is more appropriate or likely than the other. This leads to measures that are biased. Double Negatives Typical for agree-disagree questions. Please tell me whether you agree or disagree with the following starement: Employers should not make special allowances for smokers.

Cognitively more demanding to understand what it means to disagree with employers not making special allowances. Leads to unreliability as people make mistakes in interpretation.

Social Desirability Bias Over claiming positive attitudes while under claiming negative ones. People typically over report, i.e. voting in elections? Under report criminal behaviour. Under report attitudes thought to be undeseriable. i./e. under report ractial prejudice, or approval on an unpopular policitian (occurred in 1992 election polls) Options Introducing the survey question Data collection methid Use of indirect methods Minimising Social Desirability Bias Conifdentiality commitment, emphaises importance of accuracy to motivate them. Use self completion as PPs are more honest without interviewer presence. Minimise sense that answers will be negatively valued. Introduction Other people approach or casual approach: Many people find it difficult to get out and vote, maybe because they are ill or have family commitments. How about you, did you vote in the last general election? By the way, before we finish the interview, can just check to see if you happened to vote at the last general election? Measurement error in rating scales Many studies assess reliability and validity of different response formats. Reliability assessing inter-item or over-time correlation

Validity assessing correlation with other criterion variables, i.e. political interest items with education level. Number of scale points? 5 to 7 Is optimal. (5 for unipolar, 7 for bipolar) Need odd if the middle alternative is required. <5 usually too coarse >7 leads to unreliable measures as respondents do not discriminate between points

Midpoints (neither/nor)? Should they be included? Yes. If theres a genuine midpoint that makes sense i.e. attitude to capital punishment undecided? Can see pros and cons. Etc. No. Encourages Satisficing [Krosnick 1991]. Where respondents look for cues that allow them to reduce effort while appearing to answer properly. Research on this remains inconclusive. No opinion/DK No opinion filters or dont know option? If not there most people will give an opion. Some studies have asked about fictitious issues. But some people genuinely have no opinion If there, DK option allows some respondents to select this when they dont have an opinion but it can encourage satisficing. Means some may select DK when they do in fact have an opinion. Krosnick concludes that leaving out filters or DK is best, but not everyone agrees.

Using numbers and labels Labels vs numbers: Evidence that more reliable and valid data if all points are labelled. Helps to ensure constancy or meaning across respondents. Very good (1) Fairly good (2) Fairly bad (3) Very bad (4)

Choice of which verbal labels Some people asked to rate quantifiers on a scale from 0-100. Some fairly consistent findings from Krosnick-Fabrigar 1993.

Balanced Choice Well documented psychological tendency for acquiescence. People like agreeing with things (anything!) So bias introduced into survey questions that invite agreement should be avoided. Recent work has shown greater reliability and validity using balanced choice.

Unbalanced vs Balanced choice Around here, most unemployed people could find a job if they really wanted one Strongly disagree, disagree, neither agree nor disagree, agree, strongly agree Around here, do you think that most unemployed people who wanted a job would be able to find one if they really wanted or do think that they would not be able to find one? Definitely not find one, probably not find one, probably find one, definitely find one.

Questionnaire design should reduce measurement error (TSE perspective) Many sources of response variation in survey answers of which attitudes, beliefs etc are only one (true score). The aim is to design questionnaires that have as little measurement error as possible (error score).

Ordinal Variables Attributes can be rank ordered. i.e. for a variable educational attainment 1 = at least 5 GCSEs 2 = at least 2 A levels 3 = undergraduate degree 4 = postgraduate degree

The bigger the number, the higher the level of educational attainment. But distances between attributes mean nothing. No arithmetical comparisons or calculations can be applied to them. Is the difference between GCSE and A level the same as between undergraduate and postgraduate? No.

Particular numbers used as codes is arbitrary, but they must be in the correct order. Internal Variables For interval variables, distance between attributes is meaningful. Each interval is equal. i.e. temeperature. The difference between 10 degrees Celsius and 20 degrees Celsius is the same difference between 30 degrees Celsius and 40 degrees Celsius. Calculations, say, of the average temperature in London over a year, are meaningful But 20 degrees Celsius is not twice as hot as 10 degrees Celsius. Because there is no real zero point on the scale (0 degrees Celsius is a convenient but arbitrary point) we cant compare ratios.

Derivation of Mean

Calculate mean by summing all scores and dividing by number of cases. Comparing the mean and median for skewed and symmetric distributions.

QUALITATIVE
Qualitative Research Emphasises words rather than quantification in the collection and analysis of data. Its an inductive, constructionist and intepretivist research strategy. Qualitative researchers do not always subscribe to all three of these features. Naturalism Three distinct meanings. 1) a commitment to adopting the principles of natural scientific method. 2) Being true to the nature of the phenomenon being studied. 3) a style of research seeking to minimize the intrusion of artificial methods of data collection. Interpretative Thick Description Detailed accounts of a social setting that can form the basis for the creation of general statements about a culture and its significance in peoples social lives. *Geertz+ Reflexivity Reflectiveness among social researchers on the implications for the knowledge of the social world they generate of their methods, values, biases, decisions and mere presence in the very situations they investifatte. Participant Observation The researcher immerself themselves in asocial setting for an extended period of time observing behaviour, listening to what is said in conversations between others and

the fieldworker and asking questions. Usually involves interviewing key informants and studying documents. Its specifically focused on group observation. Non-Participant Observation Moderator (Facilitator) The person who guides the questioning of a focus group. Going Native Insider versus outsider issues? Objectivty vs subjectivityOpen versus closed? Overt, covert Epistemology A theory of knowledge. A stance on what should pass as acceptable knowledge. Onotology A theory of the nature of social entities. Positivism An epistemological approach advocating the use of natural science methods to study social reality and beyond. Interpretivism - An epistemological approach requiring the researcher to grasp the subjective meaning of social ce action. It respects the differences between people and objects unlike positivism. (Compare with own notes)

How and why are interpretivisim and positivism different? Onotological differences? Methodological differences? Bryman chapters 1, 2, 17 and 26. Induction An approach to the relationship between theory and research in which the former is generated out of the latter. Deduction - An approach to the relationship between theory and research in which the latter is conducted with reference to hypotheses and ideas inferred from the former. Research Question A question providing an explicit statement of what the researcher seeks. Grounded Theory An iterative approach to the analysis of qualitative data aiming to generate theory out of research data by achieving a close fit between the two. Qualitative Methods Mixed Methods Researcch Research that combines quantitative research and qualitative research. Can also describe research that combines just quantitative/qualitative research methods. Data Collection issues (ethics, sampling, recruitment, confidentiality, representation) Data Analysis A stage that features several elements P13-14 Always contrast with positivism.

WHEN TO USE FOCUS GROUPS You use them when you seek a broad understanding or overview on a subject and to clarify perceptions and definitions as PPs can respond to each other and refine their views. This allows comparing and contrasting of views and experiences.

You use them also to generate new ideas or test alternative strategies. Theyre especially good for non-complex and non-sensitive issues and subjects not too bound by social norms or conventions. Also good when PPs have something in common, i.e. similar circumstance or status.

ADVANTAGES OF USING IN DEPTH INTERVIEWS INSTEAD They allow exploration of individual attitudes, experiences or decision making and to compile detailed case studies. Its good to use them to set views in context of individual biographies or circumstances. Interviews are better for collecting complex factual material and are for sensitive issues. Much better to deal with people from very different backgrounds or statuses to avoid conflicts. The interviewer isnt a moderator and so takes a proactive approach. For following up themes in Focus groups.

Evaluating Qualitative Research Define and understand the following with application to qualitative research. Reliability Validity Generalizability Reflexivity EXAMPLES OF STUDIES TO DISCUSS Compare/contrast with quantative research.

Ethnography Researcher immerses self in a social setting for an extended period of time, observing behaviour, listening to what is said in conversations between others and the fieldworker. And asking questions. Much more inclusive than participant observation.

Problems, issues, access, gate-keepers, authorisation, ethics, safety, length study Read and relate to 2 pieces of research (at least one covert and one overt to discuss)

Data Collection methods: Interviews, focus groups, observation Film, field notes, photographs, letters, diaries

Bryman Chapter 19 Participant Observation Advantages In depth understanding Naturalistic method Oppurtunity to build trust/rapport with research subjects Open new insights into a phenomenon High Validity

Disadvantages Time consuming Negotiating access Difficult to record data (may feel awkward if done in their presence) Ethical Issues Criminal activity (Venkatesh), covert study and conent Researcher presence can change behaviour of researched

Ask clear, single questions that are truly open ended. Ask experience/behaviour questions before opinion/feeling questions. Sequence the questions? Probe and follow up questions. Interpret questions. Avoid sensitive questions. Encourage free rein but maintain control. Establish rapport. A good interviewer should be: Curious Patient Composed Resilient Attentive Flexible and quick thinking Have a good memory

Past Questions tips: Understand how to develop a discussion guide for a research project. They should illustrate clear evidence of your expertise in planning detailed discussion via appropriate structure and style of questions. Be able to explain your approach taken for the design of the discussion guide.

Discussion Establish commonality amongst PPs in the Focus Group Display evidence of thought through question order

Include issues you feel relevant to question. i.e. if its drugs, first/most recent use of drugs. Was it taken prior to coming to Uni (sample is uni students) Ask is it normal? Family/friends attitudes to drugs. Develop a coding framework after reading a section from a transcript. Analyse a document. Re. Research method questions, did the researcher use the best method available to them? Is the method a useful tool and under what circumstances?

Define the research method in the introduction and introduce a key study. Start with a question. Very often about why or how, or even what. With Research questions we typically want to explain. Sometimes we aim to describe. Thus we want theory and observation (findings). Check Bryman for other research questions (i.e. prediction, evaluation, policy creation) are variants. i.e. Why are many immigrants especially religious? Why might their children be less religious? Are young Muslims different, and why? [Voas. 2012 Islam moves West: Religious change in the first and second generations. Annual Review of Sociology 38: 525-45]

Why are many immigrants especially religious?


Social context in counties of origin Residential concentration Support from religious communities Endogamy, spousal migration, families Why might their children be less religious?

Are young Muslims different, and why?

You might also like