Quantitative Methods in Sociology PDF

Quantitative Methods in Sociology One way that sociologists collect empirical information about the social world is by using quantitative methods These are methods that either (a) measure elements of the social world that are naturally quantitative/numerical or (b) translate aspects of the soc...

Quantitative Methods in Sociology One way that sociologists collect empirical information about the social world is by using quantitative methods These are methods that either (a) measure elements of the social world that are naturally quantitative/numerical or (b) translate aspects of the social world into some quantitative measurement so that it can be analyzed using statistical methods Quantitative methods in sociology Utilizing quantitative methods in sociology is a two step process Step one: data collection Step two: data analysis Methods for Collecting Quantitative Data Original Survey The most traditional way to collect quantitative data is via a survey A survey is a questionnaire – research participants answer a series of questions that are later analyzed using statistics Research participants: The people who voluntarily agree to participate in human subject research For research projects employing original surveys, the research participants are the people who agree to answer the survey Things to consider when collecting survey data What kinds of phenomenon are we interested in? What kinds of questions should we ask? How should we find people to take our survey? How many surveys should we collect? What kinds of phenomena are we interested in? Research questions & quantitative methods Research questions generally deal with concepts Concepts are abstract terms, and are often not directly observable Example: What is the relationship between social media use, educational attainment, and mental health in North American teenagers? This research question contains three concepts: social media use, educational attainment, and mental health Variables Quantitative methods measure the social world as a series of variables A variable is a measurement of some phenomenon that has more than one value or score (i.e. that varies) In quantitative research, the concepts referred to in the research question must be developed into measurable variables Variable: University major Case: Individual person Values: categories that people can select from  Operationalization Operationalization involves specifying precisely how a concept will be measured Operationalization translates a concept into a variable or (more often) into a series of variables V1: hours per day spent on social media (time) *Link to reading: - How does Twenge (2017) V2: number of times per operationalize mental health? day someone visits a social - How does Twenge (2017) Concept: social media use media site operationalize social media use? - What other variables does Twenge (2017) consider? V3: number of social media site someone engages with Measuring social media use Questions? Who should we ask? Ok, so we’ve decided that a survey method is the most appropriate method for answering our research question The next consideration is: who should we ask to take our survey? Answering this question involves the development of a sampling procedure  Sampling procedures for quantitative research Population: the universe of cases that the research question is relevant to Sample: a subset of a population that is investigated empirically Generalizability: the extent to which observations about a sample can be reasonably assumed to represent a population Generalizability If the results of the survey are generalizable, it means that they give us a good picture of what the population looks like One of the great strengths of quantitative methods is that, under the right circumstances, they can give us a good picture of some social phenomenon even if we only study a small part of that phenomenon empirically The two factors that influence generalizability are: The sampling procedure The sample size à The larger the sample size, the more likely the results are generalizable Sampling procedures and generalizability Random Each individual in the population has an equal BETTER probability of being selected for study Representative The sample is a reproduction of the population along particular demographic characteristics Convenient People are sampled based on their availability Snowball sampling People that have been sampled introduce the researcher to other possible study participants -Usually limited to qualitative research WORSE -Often the only way to sample difficult to access groups Questions? Other quantitative methods in sociology Secondary analysis Data scraping Quantitative content analysis Secondary analysis When researchers use secondary analysis, they analyze existing data in a novel way (rather than collecting original data) Census, GGS, CCHS, etc. Advantages Sample size Sampling technique Cost Data scraping Using computer algorithms to generate data about people’s behaviour by “scraping” information about their online activity Procedure that originated in marketing In sociology, a useful tool for overcoming social desirability bias Social desirability bias Social desirability bias describes the fact that people may answer questions on a survey based on how they wish to appear, rather than how they actually behave Can be conscious or unconscious Example: GGS: How often do you attend a faith institution? 18% of respondents say they attend a faith institution once per week Results from cellphone data analysis: 3% Quantitative content analysis Content analysis is the systematic analysis of media that we consume Quantitative content analysis involves counting how often something occurs within a set of textual, visual, or auditory media Quantitative Analysis What kinds of variables are we working with? Variable type Definition Example Nominal / Categorical Numbers are used to represent different Race, Neighborhood, Marital conditions, but the phenomenon itself is not Status, Religion, Favourite quantitative – therefore, the variable values Kardashian cannot be ranked Ordinal Different values of the variable can be ranked, Likert scales, SES/Class, Pain but there is no way to measure the precise difference between ranked values Ratio Differences between values are measurable, Number of siblings, Income, Hours and there exists a real zero (limit) spend on social media per day Nominal variables Ordinal Variables Ordinal variables: Likert scales Likert Scale: A Likert scale is a survey question that measures survey participants’ opinions, attitudes, or motivations. The variable values of a Likert scale quantify some phenomena (e.g. “level of agreement”) that is not naturally quantitative. Ratio Variables Ordinal vs. Ratio variables It is important to note that the variable type is determined by how a variable is operationalized Questions? Descriptive Statistics Descriptive statistics tell us about the distribution of one variable This makes descriptive statistics univariate statistics Central tendency: measures of central tendency attempt to give a quick picture of the content of one variable Mode The variable value that is the most common, or has the highest count For nominal level variables, the mode is the only appropriate measure of central tendency Median The value that separates the sample into two equal halves The “middle value” !"# #%$"# To find the middle value: $ ; $ = 51.5 N = 102 Middle value = 51.5 Median variable value = 3 or less  Mean The average value ∑ "! 𝑥̅ = # à sum of variable values / n (number of cases) N = 114 Average number of siblings: 2.06 Outliers Outliers are extreme cases (variable value is extreme relative to the majority of the distribution) Outliers overinfluence the mean As such, it is sometimes appropriate to remove outliers from analysis Mean when outlier is included: 2.06 Mean when outlier is excluded: 1.95 Measures of central tendency and distribution skew Variable distribution: The way a variable is distributed across its values For ratio variables, the measures of central tendency can tell us something about how the variables is distributed Descriptive statistics Proportion: Tells us the percentage of a variable that falls into one particular variable value Related as a value between 0 and 1 (Twenge, p. 60) “[In] 2015, one out of three high school seniors admitted they had not read any books for pleasure in the past year.” 1 out of 3 = 0.33 àThe proportion of high school seniors who did not read for pleasure in the past year is 0.33 àRelated as a percentage: 33%  Class data: proportion of students who did not read any books for pleasure last year = 0.245 # $% &'()( *+ ,-) &',).$/0 $% *+,)/)(, 12 To calculate the proportion: → = 0.2105 + 332  Class data: proportion of students who did not read any books for pleasure last year = 0.245 percentage proportion # $% &'()( *+ ,-) &',).$/0 $% *+,)/)(, 12 To calculate the proportion: → = 0.2105 + 332 Questions? Inferential Statistics Inferential statistics measure the relationship between two or more variables Knowing the value of one variable allows us to make an inference about the likely value of another variable Quantitative methods in sociology make inferences about the relationship between variables in the population, based on inferential statistics calculated using data from a sample Inferential statistics can be: Bivariate statistics: describe the relationship between two variables Multivariate statistics: describe the relationship between three or more variable Independent vs. dependent variables Independent variable: The variable that is hypothesized to influence the dependent variable Dependent variable: The variable that is hypothesized to be influenced by the independent variable Independent Dependent variable variable Hours spent Number of depression risk on social factors someone media has Independent vs. dependent variables Which is which? Sometimes, there is a logical reason to consider one variable to be the independent variable, and the other to be the dependent variable For example, if we want to study the relationship between gender and income, we will consider gender to be the independent variable (because gender is logically prior to income) In many situations, however, is it not clear which variable is the “influencer” and which variable is the “influenced” Example: Does social media use cause depression, or are those who are depressed more likely to use greater amounts of social media? Inferential statistics and theorizing causation Inferential statistics can tell us the extent to which two or more variables share a mathematical relationship For example, if a correlation coefficient shows that two variables, x and y, are related, then we know that a change in variable x gives us some predictive capacity with regard to the value of variable y (and visa versa) But the correlation coefficient does not tell us about causation This means that we need to theorize the relationship: 1. x causes y 2. y causes x 3. the relationship between x and y is spurious Example: incarceration and probability of TBI Multivariate statistics In sociology, quantitative research almost always uses multivariate analyses Multivariate statistics describe the effect of several independent variables at once, on some dependent variable We need multivariate statistics because we do not use experiments (lack of random assignment) Multivariate analyses usually employ control variables: variables that are not directly related to the research question but are suspected to be related to the dependent variable In sociology, common control variables include gender, race, and social class Dependent variable: Attitude towards homosexuality Independent variables: (1) GDP per capita (2) occupational category Control variables (not visible in this graph, but included in analysis): Age, education, martial status, religion, town size, religious composition of country