SOC101Y Introduction to Sociology Professor Adam Green Professor Robert Brym Lectures #20 & 21 Research Methods 27 Mar & 3 Apr 13

OTTFFSSENT O{ne}T{wo}T{hree}F{our}, etc. ADGJM A{BC}D{EF}G{HI}J{KL}M

How Research Filters Perception Values, Theories, Existing Research, Methods REALITYREALITY

The Research Cycle 1.Figure out what matters to you 2. Formulate a testable theory (a tentative explanation of a phenomenon) 3. Review existing literature 4. Select method(s) 6. Treat subjects ethically 5. Collect data 7. Analyze data 8. Report results

Research Ethics 1.Respect your subjects’ right to safety. Do your subjects no harm and, in particular, give them the right to decide whether and how they can be studied. 2.Respect your subjects’ right to informed consent. Tell subjects how the information they supply will be used and allow them to judge the degree of personal risk involved in supplying it. 3.Respect your subjects’ right to privacy. Allow subjects the right to decide whether and how the information they supply may be revealed to the public. 4.Respect your subjects’ right to confidentiality. Refrain from using information in a way that allows it to be traced to a particular subject. 5.Do not falsify data. Report findings as they are, not as you would like them to be. 6.Do not plagiarize. Explicitly identify, credit, and reference authors when making use of their written work in any form, including Web postings.

Participant Observation  Researchers engage in participant observation when they try to observe a social milieu from an outsider’s point of view and take part in the activities of their subjects (allowing them to understand the point of view of their subjects).  They must avoid influencing their subjects’ behaviour (reactivity or the Hawthorne effect).  Most participant-observation studies begin as exploratory research in which the researcher uses hunches as hypotheses (unverified but testable statements derived from theories).

Measurement  Variables are concepts that can take more than one value.  Operationalization involves establishing criteria for assigning values to variables.  If a measurement procedure yields consistent results, we consider it reliable.  If a measurement procedure measures what it is supposed to, we consider it valid (and it has external validity if it is consistent with what we know from previous research or allows us to make useful predictions).  If research findings hold in many contexts, we consider them generalizable.  Causality is the measurement of causes and their effects.

Measurement as Target Practice x xx Validity, reliability, and generalizability may be explained by drawing an analogy between measuring a variable and firing at a bull’s-eye. In case 1, above, shots (measures) are far apart (not reliable) and far from the bull’s-eye (not valid). In case 2, shots are close to each other (reliable) but far from the bull’s-eye (not valid). In case 3, shots are close to the bull’s-eye (valid) and close to each other (reliable). In case 4, we use a second target. Our shots are again close to each other (reliable) and close to the bull’s-eye (valid). Because our measures were valid and reliable for both the first and second targets in cases 3 and 4, we conclude our results are generalizable. x xx 1.Not Valid, Not Reliable 2. Not Valid, Reliable 3. Valid, Reliable 4. Valid, Reliable, Generalizable (Target 2)

Sampling  A sample is part of a group.  A population is the entire group.  A voluntary response sample is a group of people who chose themselves in response to a general appeal.  A representative sample is a group is a group of people chosen so their characteristics closely match those of the population of interest.  A convenience sample consists of people who are easiest to reach.  If respondents are chosen at random and an individual’s chance of being chosen is known and greater than zero, the respondents constitute a probability sample.  A sampling frame is a list of all the people in the population of interest.  A randomizing method is a way of ensuring every person in the sampling frame has a known, equal, and non-zero chance of being selected.

Sampling Error 48% Conservatives { X } % margin of error 50% Liberals { X } % margin of error Because the sample measures fall within overlapping margins of error, we conclude that the measured difference in the popularity of the two parties is not statistically significant.

Surveys  A mail questionnaire is a form containing questions is mailed to the respondent and returned to the researcher through the mail system.  The response rate is the number of people who answer the questionnaire divided by the number of people asked to do so, expressed as a percent.  In a face-to-face interview survey, questions are presented to the respondent by the interviewer during a meeting.  In a telephone survey, questions are presented to the respondent by the interviewer over the phone.  A closed-ended question provides the respondent with a list of permitted answers.  Open-ended questions allow respondents to answer questions in their own words.

Threats to Validity  Undercounting occurs due to an imperfect sampling frame.  Nonresponse occurs when respondents do not answer some or all questions.  Response bias occurs when respondents do not answer questions completely accurately.  To avoid wording effects, questions should be specific, simple and neutral, and they should focus on important, singular, current events.

more than 10 hours TV per week and no act of physical violence per year 10 or fewer hours TV per week and at least 1 act of physical violence per year 10 or fewer hours TV per week and no act of physical violence per year Turning a Classroom into a Contingency Table (a cross-classification of cases by at least two variables) more than 10 hours TV per week and at least 1 act of physical violence per year BACK LEFT FRONT RIGHT

TV Viewing By Aggressiveness (in percent) Measures strength of association Number of cases in each column Percent of cases in each column Percent of cases in each column that fall into a category of each row variable. Step 1: Note the direction in which the table is percentaged. C O L ROW U M N

TV Viewing By Aggressiveness (men only)

Testing an Association for Spuriousness (1) We believe there is a causal relationship between TV viewing and aggressiveness: TV viewing Aggressiveness (independent variable)(dependent variable) (2) By controlling for gender we can see whether gender has created a spurious association between TV viewing and aggressiveness: TV viewing (independent variable) Respondent’s gender (control variable) aggressiveness (dependent variable) (association) (no association)

Find the control variables in the next five slides:

Female/male earnings ratio Year and Status Female/Male Earnings Ratio, Canada, Full-time ratio = 1 in 2079, when today’s 19-year-old is 92 years old.

Nuclear Family Decline: USA and Sweden, Early 1990s USA Sweden median age at first marriage men * women * percentage of population never married men * women * nonmarital birth rate * 1-parent hshlds among all hshlds with children < *18.0 % of mothers in labor force with children < * total fertility rate average household size *

Child Well-Being: USA and Sweden, Early 1990s USA Sweden mean reading performance score at * % of children in poverty single-mother households * two-parent households * death rate of infants from abuse * suicide rate for children (/100,000) * juvenile delinquency rate (/100,000) 11.6*12.0 juvenile drug offence rate (/100,000) *

Number of Sex Partners by Respondent’s Sex, USA, 2002 (in %) respondent’s sex malefemale number of sex partners 0 or more than total n 1,004 1,233

Number of Sex Partners by Respondent’s Sex, USA, 2002, Married People Only (in %) respondent’s sexmalefemale number of sex partners 0 or more than total n

Regression Analysis Y X x xx x xxx x x y = a +bx, a is the value of the intercept, b is the value of the slope, y is the value of the dependent variable, x is the value of the independent variable

A Linear Regression Example 7. The correlation coefficient (r=0.84), says the diamonds are not highly scattered around an upward sloping regression line: a strong, positive correlation exists between age and income. 6. The regression equation (y=a+bx), says that the predicted value of the dependent variable (y) at a given value of the independent variable (x) equals the intercept (a= ) plus the value of the independent variable times the regression coefficient (1.0876). 4. The intercept (a) is the value of y at the point where the regression line cuts the vertical axis. 5. The slope of the regression line (b; the regression coefficient) equals the “rise over the run,” that is, the change in the dependent variable (y) for every unit change (here, 1 year) in the independent variable (x). Here, the regression coefficient = , meaning that that for every 1-year change in age, income changes by x $1,000, or $1, Each diamond represents an individual’s age and income. The regression line minimizes the square of the sum of the vertical distances between each diamond and the line. The regression line predicts the value of y for a given value of x. 1. The cause or independent variable (x). 2. The effect or dependent variable (y). y = x r =.84

Correlation r =.85 r = -.92 r = 0 Dependent variable Dependent variable Dependent variable Independent Variable Independent Variable Independent Variable 1. Positive Correlation 2. Negative Correlation 3. No Correlation