SOCI 373 Lecture Notes - Lecture 6: Content Analysis, Big Data
19 Jun 2018
School
Department
Course
Professor
02/06/2018
Methods cont.
Experiment is considered the gold standard for causality.
Cross sectional surveys- examining people at one point in time. Pros- you don’t lose people,
don't have to worry about mortality affecting results. Cons- Even if there is an association at one
point in time, how do you know that this isn't reverse causation?
Longitudinal Research- done over time. Includes: (1) Repeated Cross Sectional(you still can't
establish causality) same survey given to different groups of people repeatedly, useful for a
trend analysis. Change over time in the general population, you don't have to worry about
survey attrition. You still cannot determine causality. Most often how individuals are studied in
Canada.
(2) Panel Design- a second type of longitudinal design. Allows more influence on causality.
Follows the same people over time. Then you have to worry about survey attrition, fatigue,
familiarity. You have a panel of people from only one time, not the upcoming generation. One
way of addressing this is to use a refresher study, add in additional people over time(a new
cohort).
Canadian Longitudinal Study of Aging- just began, only first set of observations are available, so
right now it is only a cross sectional.
Secondary data analysis- the researcher does not gather data, instead existing data is used.
Typically this is from Stats Canada. Cons- you are at the mercy of what other people want to
ask ex governments don't want to ask about sex lives of older adults.
P-hacking: someone gets a big data set that has already been gathered, and just by chance
finds an association without theory guiding it. Does not mean that the results could be
replicated.
Pros: it takes less time and money than conducting your own research. You must be honest in
your research results, anyone can analyze your data. Secondary data analysis can facilitate
interdisciplinary research.
Seniors and Surveys:
It may be difficult to get older adults to take your survey, they may be more likely to drop out or
die during your study.
You must take mortality (sample selection) into account.
Older adults are more likely to respond with ‘response sets’ ex yay saying and nay saying
(agreeing or disagreeing with everything)
Older adults may be less willing to respond to questions about sensitive topics.
Secondary data analysis can also be performed on textual materials ex content analysis of
news stories, how are news sources presenting aging in the media. More qualitative than
quantitative.
Historical and Literary Methods- involves detailed interviews with older respondents about their
life course. Affected by bias and retrospection.
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Experiment is considered the gold standard for causality. Cross sectional surveys- examining people at one point in time. Pros- you don"t lose people, don"t have to worry about mortality affecting results. Includes: (1) repeated cross sectional(you still can"t establish causality) same survey given to different groups of people repeatedly, useful for a trend analysis. Change over time in the general population, you don"t have to worry about survey attrition. Canada. (2) panel design- a second type of longitudinal design. Then you have to worry about survey attrition, fatigue, familiarity. You have a panel of people from only one time, not the upcoming generation. One way of addressing this is to use a refresher study, add in additional people over time(a new cohort). Canadian longitudinal study of aging- just began, only first set of observations are available, so right now it is only a cross sectional. Secondary data analysis- the researcher does not gather data, instead existing data is used.
