Research Flashcards
Normative analysis
prescriptive, based in reason and logic
Empirical analysis
descriptive / explanatory, based on observation and measurement
Scientific approach
an understanding of knowledge (epistemology) and way of obtaining knowledge (methodology)
Positivism
there is an objective reality
Interpretivism
reality changes with perspective and is decided by the individual
Core beliefs of scientific approach
- Empiricism: knowledge is derived from real world observation, not theoretical deduction
- Determinism: everything has a cause that we can find
- Objectivity: science should accurately represent reality
- Replication: science is cumulative, so we need to repeat research to make sure it’s correct
Intersubjectivity
multiple studies should demonstrate similar findings
Components of a research report
- Abstract/executive summary
- Introduction
- Research design
- Presentation of findings
- Discussion
- Conclusions
- References
Null findings
research that results in no proven connection between 2 concpets
Replicable
showing you research so others can replicate it to prove your hypothesis
Transmissible
using research that is easy to understand
Informed consent
research participants need to fully understand the extent to which they will participation and the nature of the project
Systematic errors
getting the same wrong answer after multiple attempts
Random errors
getting different answers each time you attempt
Levels of measurement (NOIR)
- Nominal: only names and categories
- Ordinal: ordered information
- Interval / Ratio: exact numbers
Applied research
research done to solve a real world problem
Basic research
research done for the sake of understanding new ideas
Measurement errors
gap in between you expected result and the actual result
Positive correlation
both variables move in the same direction (up or down)
Negative correlation
variables move in opposite directions
Causation
one concept happens because of another concept
Correlation
2 concepts move similarly, but this may not be because of each other
Independent variable
the variable that isn’t changed by the other (cause)
Dependent variable
the variable that changes based on the other one (effect)
Types of random sampling
- Simple: draw from a group
- Systemic: creating an algorithm to draw from a group
- Proportionate: random selection based on the percentages in the population
- Disproportionate: weighted random selection to over-represent certain groups
Types of non-random sampling
- Convenience / Accidental: sampling the first people you can find
- Volunteer: people select to be a part of the sample
- Purposive / Judgemental: manually creating a sample from personal judgement and knowledge
- Snowball: starting with a small group who then ask others like them
- Quota: selecting people to fill a requested list
Sampling error
difference in sample statistic and population parameter
Margin of error / confidence interval
how close the sample is to the population (as represented by a % range)
Document analysis
gathering key facts and basic level information from documents (names, dates, times, numbers)
Text analysis
systematic study of content, form, and substance of a text
Discourse analysis
what the text says about society, meaning, and interactions
Content analysis
message of the text, frequency of terms, length of text
Structural content analysis
physical measurements of the content (pictures, titles)
Substantive content analysis
what is being said or written (idealogical leanings, coverage bias)
Manifest content
visible surface content
Latent content
underlying meaning of the content
Codebook
rules of the analysis (what words you are looking for, what part of the content)
Intercoder reliabilty
multiple people reading or watching the same content and getting the same result
Interviewer effect
the impact the interviewer’s presence has on the interviewee and the information shared
Observation research
observing behaviour in it’s natural setting as it occurs
Participant observation
being a part of the group you are studying
Obtrusive participant observation
the group knows they are being studied
Unobtrusive participant observation
the group does not know they are being studied
Covert participant observation
undercover observation of a group (illegal)
True participant observation
researching a group you already belong to
Participant observer
identifying yourself to the group you will be studying
Complete observer
observing the behaviour of a group from outside the group
Reactivity
the researcher’s impact upon the group they are studying and the information shared
Data saturation
no new information is gained from new interviews and studies
Interview framework
guidelines of what information you want to ask in the interview
Hawthorne effect
people will change their answers when they know they are being studied (social desirability)
Focus group
small group of similar people with the purpose of gaining new and specialized information
Secondary data
data collected by another researcher that you are allowed to use
Closed-ended questions
participants have to choose from a preset group of answers
Open-ended questions
participants can answer in any way they want
Exhaustive
everyone taking a survey has an option for them (or “other”)
Aggregate data
all the data collected combined to create an average
Microdata
the individual data collected from each case
Metadata
technical information about how the information was collected
Omnibus survey
multiple groups run one survey and get relevant information for them from it (to save money)
Cross-sectional study
one-time question asked at one moment in time
Longitudinal / Panel study
long-term study that asks multiple questions about a topic over time
Non-response bias
people who agree to take the study may not accurately represent the population
Research design
imposed controlled restrictions with the purpose of observation
Experimental / Treatment group
the group given treatment before the post-test
Control group
the group not given treatment (or given placebo) before the post-test
Between-subjects design
doing the experiment without giving a pre-test
Within-subjects design
giving both groups a pre-test to better contextualize the results of the post-test
Internal validity
controlled space where you can be sure the outcome is a direct result of the treatment
External validity
real world where the outside factors may influence the outcome
Quasi-experiment
looking at statistics from the past to conduct an experiment as if it was happening now
Factorial design
testing multiple factors, the sub categories within those factors, and how they relate to each other
Double-blind design
both the participant and the administrator do not know if they are in the treatment group or the control group
Single-blind design
only the participant does not know if they are in the treatment group or the control group
Small-N research
qualitative approach, usually less than 30 cases
Large-N research
quantitative approach, large number of cases
Case study
in-depth investigation of a single individual, group, or event
Descriptive case study
great detail of everything that happens in a case
Theory-testing case study
cases that confirm of dispute current theories
Failed most-likely case
case outcome is expected to confirm a theory, but it disproves it
Successful least-likely case
case outcome is expected to disprove a theory, but it confirms it
Process tracing
explaining each step of a case development to demonstrate the causation between 2 concepts
Comparative research
small-N, contrasts cases to strengthen generalizations
Most-similar-systems design
cases with the same factors that get different outcomes, so causation can not be proven
Most-different-systems design
cases with different factors that get the same outcome, so causation can not be proven
Galton’s problem
2 different things under observation may influence each other and lose their independence
Audit trail
detailed description of the research steps taken
Cross-tabulations
table used to compare 2 nominal or ordinal variables
- IV: row (across), DV: column (down)
Descriptive statistics
statistics used to quantitatively describe information
Inferential statistics
statics used to infer from a random probability sample to a population
Univariate statistics
describes or infers a relationship between the value of 1 variable
Bivariate statistics
describes or infers a relationship between the values of 2 variables
Multivariate statistics
describes or infers a relationship between the value of 3 or more variables
Frequency distribution
how many cases take each value (raw and relative)
Raw frequency distribution
exact number of cases in each value
Relative frequency distribution
proportion of cases in each value represented by a %
Measure of central tendency
the most typical number (one number that represents the entire distribution)
Measure of dispersion
how much the values vary
Measures for nominal variables
- Central tendency: mode (most) (value with greatest number of cases or highest %)
- Dispersion: variation ratio (% of everything not in the modal category)
- Association: Lambda (PRE-based), Cramer’s V (not PRE-based)
Measures for ordinal variables
- Central tendency: median (middle) (value of the middle case)
- Dispersion: range (value of the lowest category to value of the highest category)
- Association: Gamma (PRE-based), Tau-B (symmetrical table), Tau-C (asymmetrical table)
Measures for Interval Ratio variables
- Central tendency: mean (average) (average of all values, add values then divide by # of cases)
- Dispersion: Standard deviation (average of all dispersions from the mean)
- Association: Pearson’s R (linear data), Spearman’s RHO (non-linear data)
Standardized scores
exact number of standard deviation units a case is above or below the mean
Proportional Reduction in Error (PRE)
how much knowing the value of the cases in the IV helps you predict the values of the DV
Basic linear regerssion
the regression line crosses the graph at the closest value to every point
Intercept line
what number the regression line starts on the graph
Type 1 error
assume that2 variables are related, but they are not
Type 2 error
assume that 2 variables are not related, but they are
Chi-Square
gives the likelihood of each possible degree of relationship occurring in a sample if there was no relationship in the population (nominal and ordinal only)
Control variable
the variable that is unchanged to better understand the relationship between variables
Statistical significance
how likely it is that a relationship between 2 variables in a sample might have occurred by chance and may not exist in the population
Difference of means
- T-Test: compares the means of 2 cases
- ANOVA: compares the means of multiple cases