Search
Browse
Create
Log in
Sign up
Log in
Sign up
Upgrade to remove ads
Only $2.99/month
Midterm 2
STUDY
Flashcards
Learn
Write
Spell
Test
PLAY
Match
Gravity
Terms in this set (49)
4 hurdles to establishing causation
1. is there a credible causal mechanism
2. can we rule out reverse causation
3. is there a correlation between x and y
4. have we controlled for all confounding variables
how are quantitative methods helpful for?
1. establishing correlations
2. controlling for confounders
(if designed well, it can also help study causal mechanisms)
Quantitative research process steps
1. what is your unit of analysis
2. control variables
3. getting and validating data
4. statistical techniques and programs (very basic)
5. common problems to avoid
6. interpreting statistical results
define unit of analysis
the who that is being studies; the subjects of our study
observation in a large-N analysis
one data point, or one row in your dataset; the # of observations depends on your unit of analysis
Examples of unit of analysis
countries, individuals, schools, states, cities, police departments, year
Unit of analysis over time
sometime unit of analysis includes the thing we're studying and an element of time; particularly when we are examing multiple subjects over time; then our UoA could be subject in each time period (examples: country-year; person-year; state-year)
Unit of analysis depends on
how you frame the question or hypothesis; sometimes one theory has different observable implications/hypotheses, each with different units of analysis
Why is it so important to figure out what an observation will be in your study?
it determines what data you collect
define domain of your analysis
the spatial and temporal scope; spatial is what cases will be included in the study (individuals, countries, states, leaders, Congress members, etc.); temporal: over what time period
How to decide what the domain is?
theory/logic (ex. Why do US women vote differently than men? => spatial domain is the US; temporal domain is the time after US women got right to vote)
practicality (data may be unavailable or too difficult to collect for some time periods)
policy importance
ultimately the researcher can choose the domain and explain why it's appropriate
population
full set of relevant cases
Once you've chosen unit of analysis and domain, you know what?
what your population is
descriptive statistics
a quantitative description of your data that organizes and summarizes data to make it more intelligble (average, median, percentage, rate, standard deviation)
inferential statistics
procedures for determining the extend to which one may generalize beyond the data (typically used for hypothesis testing)
histogram
graph in which the height is a vertical bar represents the frequency or percentage of cases in each category of an interval/ratio variable
bivariate analysis
the relationship between your main IV and your main DV => not control variables; the type of test depends on the level of measurement of your variables (DV: ordinal, IV: ordinal)
p-values
measure of statistical significance, given size of coefficient and number of observations; look for a p-value of < 0.05
Problems with large-N analysis
reverse causation and spurious correlations due to omitted variable bias
main goals of statistical analysis
establish whether and how two or more variables are associated (bivariate analysis, multiple regession so you can control for potential confounders); establish how certain you are that the association did not rise by chance alone (p-values, statistical significance)
qualitative research
research that is done on a small number of observations (small-n)
Strengths of case studies
ruling out deterministic theories; inductive theory building;
inductive theory building
1. take a research question you are interested in
2. identify cases that would help you develop some hypotheses about what IVs cause DVs and why (case selection that is typic and generalizable)
3. then you need new data to test the resulting hypotheses
problems with small-n case studies
tricky to test a hypothesis because hard to rule out alternative causes when you have so few causes; hard to gerneralize; counterexamples
small-n research is
good at assessing whether a causal mechanism operated in certain cases and at ruling out reverse causation; bad at establishing convincing correlation and ruling out confounding variables
two complementary ways of using evidence
comparisons across cases; analysis within a single case (process tracing)
various hypothesis testing techniques
content analysis, case control/controlled comparison method, process tracing
2 main ways of thinking about evidence
1. comparisons across cases (case control/controlled comparison/comparative method) - selected a small number of cases and reach conclusions by comparing across them
2. process tracing (select a small number of cases and take each one on its own terms)
qualitative research is good for
1. inductive theory-building
2. ruling out deterministic theories (though these are rare)
3. testing theories when the N is small
4. complementing last N analysis, especially looking for causal mechansims
5. and testing theories when measurement is difficult
selecting cases for comparaive analysis
look at the hurdles to establishing causation - big challenge with small-N is to rule out confounding and to establish correlation; a way to do this is to compare cases that have the same values on potential confounding variables (if the values on the confounders are the same, then they do not predict variation in the DV)
process tracing technique
helpful for probing causality within a case; divide the cause-effect link that connects IV and outcome into smaller steps then look for evidence of each step; think of this as deriving as many observable implications of your theory as possible
counterfactual reasoning
if X had led to Y, what would we have observed?
in order to rule out confounding and establish a convincing correlation with a small-N
compare cases that have the same values on potential confounders; think about as many observable implications as you can, including implications about causal mechanisms; after deducing observable implications, gather evidence so that you can trace the proposed causal process
Archival research
researching collections of original unpublished material (primary sources)
Challeneges when using archives
1. access to collections
2. incomplete records
3. redaction of existing records
4. interpretation of info
Trachtenberg's Method
method of learning about history by critically analyzing secondary sources (and eventually moving to primary); figure out what the most important secondary books are on a topic; read those books and think about logic, evidence; go to original source and think about whether they are being interpreted appropriately
Field Research
when you can't get info from locally available sources, you need to go elsewhere
Techniques used: interviews and surveys, participant observation, direct observation, collection of written materials
Interviews and surveys
unstructured interviews and structured interviews (with or without closed ended questions); need to using nonprobability sampling to find people to interview because of practicality; ask non threatening questions, be flexible with questions, pay attention to what is said between questions, tape and transcribe the interviews
snowball technique
ask those already interviewed to suggest the names of potential interview subjects
Problems with interviews
1. interview samples are too small to detect subtle effects with precision
2. end up with a lot of variables in open-ended questions (your N needs to be greater than the number of variables you study)
3. samples are often non representative (selection bias)
4. problems with intercodal reliability/ different interviews get different responses
5. no way to tell if people are truthful
selection bias
occurs when we analyze data that contains data subject to selection effects which leads to biased conclusions because biased sample
participant observation
researcher becomes participant in the culture or context being observed; tricky because you have to enter the context without chaging it with your presence and takes years of work
Advantages to participant observation
gain access to what you wouldn't have before; opportunities for induction
Disadvantages to participant observation
presence may alter the studies phenomenon, subjective method, selection bias
Direct observation
try to unobtrusively observe a phenomenon (take video, observe protests, public meetings)
Advantages to direct observation
presence is less likely to affect outcome
Disadvantages to direct observation
may or may not get unbiased info, subjective, selection bias
collection of written materials
archives, pamphlets, flyers, etc.
Unit of analysis
the thing that constitues a single observation in your study; figure out what the entity and interval are that make up a single observation in your study
YOU MIGHT ALSO LIKE...
CCJ 4700 Final Exam
34 terms
Sociology 301 Final
77 terms
Research Chapter 17: Analyzing Existing Data - Qua…
35 terms
Secondary Data Analysis and Content Analysis
10 terms
OTHER SETS BY THIS CREATOR
FINAL
90 terms
Final
68 terms
Midterm 1
29 terms
Brazil v Mexico
4 terms