Test 1 Flashcards
Normative Statements
Arise from value judgements and based on societal values. Can not be tested or proven True or False.
Remember: “Opinion-Based”
Example: “A higher voter turnout would make democracy stronger.”
Empirical Statements
Statements based on facts, observable and measurable evidence. Can be tested and verified through research and data collection.
Remember: “Evidence-Based.”
Example: “Voter turnout in the 2020 U.S presidential election was approximately 66%.”
Prescriptive Statements
Recommendation on how to achieve a desired goal. These statements combine empirical and normative elements by recommending specific actions based on normative beliefs.
Remember: “Solution-Based”
Example: “The government should make election day a national holiday to increase voter turnout.”
Faith / Appeal to Authority
Knowledge based on trust in authority sources such as religious texts, traditions, or experts.
Remember: Relies on Belief
Example: “We should adopt this policy because a supreme court justice said so.”
Reason
Knowledge derived from logical thinking; makes sense; role of assumption
Remember: Rational thought and evidence
Example: “If free trade increases economic growth and economic growth reduces poverty, then free trade should reduce poverty.”
EROS - (Every Researcher Observes Science)
Empiricism
Knowledge is based on observation and experience, not intuition or faith.
Example: Political Scientist study voting behavior by analyzing real election data, not just personal opinions.
EROS - (Every Researcher Observes Science)
Replicability
Research should be repeatable by others to confirm findings.
Example: If a study finds that negative ads decrease voter turnout, other researchers should be able to repeat it and get similar results.
EROS - (Every Researcher Observes Science)
Objectivity
Findings should be free from personal bias and based on facts.
Example: A researcher studying campaign finance should analyze data impartially rather than letting political beliefs influence results.
EROS - (Every Researcher Observes Science)
Skepticism / Falsifiability
Scientist should question claims and demand evidence before accepting conclusions. Theories must be testable and capable of proven wrong.
Example: A claim that social media increases voter turnout should be examined critically and tested with data before being accepted.
Theories
A set of related propositions which attempt to explain why events and relationships occur. They are usually tentative formulations.
Fact: In hard sciences, theories could be laws
Remember: (THEO- The Helpful Explanation of Outcomes”
Example: “Democratic peace theory suggests that democracies are less likely to go to war with each other.”
Hypothesis
More specific statements that arise from theories. It is important to realize that hypothesis derive from theory. Sometimes Hypothesis and theories are one in the same.
Remember: The real difference between theory and hypothesis is that hypothesis is usually more specific and more testable.
Example: “Democracies are more likely to engage in peaceful negotiations with other democracies.”
Four Aspects of Hypothesis
Certainty Factor: How confident we are about expected relationship
Causation V Association: Does one variable cause the other or are they just correlated?
Direction: Is this relationship positive or negative?
Number of Components: How many variables are invovled?
Dependent Variable (Y)
The phenomenon one is trying to explain or predict.
Remember: Depends on the Independent Variable.
Independent Variable (X)
The presumed cause of change in the dependent variable.
Remember: The independent variable influences the dependent variable.
Operationalization
Defining how a variable will be measured using specific indicators.
Nominal Scale
Categories without a meaningful order
Example: Political party affiliations)
Ordinal Scale
Categories with an order but no fixed intervals
Example: Military Ranks
Interval Scale
Ordered categories with equal intervals
Example: Temperature
Types of Surveys
Mail
Cheap, but low responsive rate
Types of Surveys
Phone
Allows clarification but excludes non-phone users
Types of Surveys
Face to Face
High Response Rate but Costly
Types of Sampling
Random Sample
Everyone has an equal chance of selection
Types of Sampling
Stranded Sample
Population divided into subgroups, then randomly sampled
Types of Surveys
Group Administered
Ensures high response rates and control over the survey process
Types of Surveys
(CATI) Computer-Assisted Telephone Interviewing
Increases efficiency, reduces errors, and allows for standardized questioning.
Types of Surveys
(CAPI) Computer-Assisted Personal Interviewing
Improves accuracy and allows for real time data collection while maintaining personal interaction
Types of Surveys
Exit Polls
Helps predict election results before official counts and analyzes voter behavior
Types of Surveys
Web
Cost-effective, allows for large scale data collection and provides convivence for responders.
Sampling Error
Errors that arise due to the fact a sample is being used instead of the entire population.
Non-Sampling Error
Errors that occur during the data collection of analysis process, unrelated to sampling method.
Non- Sampling Error Type
Measurement Error
Occurs when there is a problem with how questions are asked, how data is recorded, or how the respondent understands the question.
Non-Sampling Error Type
Nonresponsive Error
When some individuals in the sample do not respond. This can lead to bias results if the non-respondents differ in important ways from respondents.
Types of Sampling
Stratified
Ensures that each subgroup is represented proportionally, especially when there are significant differences between strata.
Types of Sampling
Disproportionate
Useful for ensuring that small or underrepresented groups have adequate representation, thought it may require weighting the results to correct the disproportionately.
Internal Validity
The degree to which an experiment or study accurately measures the causal relationship between variables, free from the influence of external or confounding factors.
Example: In a study on voter turnout, internal validity would refer to whether the observed changes in turnout are truly caused by the intervention (e.g., a new voting law) rather than other factors (e.g., socioeconomic status).
External Validity
The extent to which the results of a study can be generalized to other settings, populations, or times.
Example: A study on voting behavior conducted in a particular state should have external validity if its findings apply to other states or national elections.
Construct Validity
The extent to which a test or survey measures the concept or theory it is intended to measure.
Example: If a survey aims to measure political engagement, its construct validity would depend on whether the questions truly capture the concept of engagement (e.g., voting, attending political events, etc.) rather than irrelevant behaviors.
Content Validity
The degree to which a measure fully represents the concept it’s intended to assess. This usually involves experts evaluating whether the questions or items in a test comprehensively cover the entire domain of the concept.
Example: A political science exam should have content validity if it covers all major topics (e.g., theories of democracy, elections, policy analysis) rather than focusing too much on one narrow area.
Criterion-Related Validity
The extent to which a measure is correlated with a relevant outcome or criterion that it should theoretically predict. Criterion-related validity can be further divided into.
Example: A job performance test in political campaigns may have predictive validity if it can successfully predict the future performance of campaign staff.
Issues of Reliability
Inconsistency, Measurement Error, Observer or Rater Bias, Time Variability, Contextual Factors
Interrelation of Statements
Empirical, normative, and prescriptive statements interrelate by forming a logical progression in research and analysis. First, empirical statements provide the factual basis of what is happening in the world. Then, normative statements offer a value-based interpretation of those facts, suggesting what should happen or what is ideal. Finally, prescriptive statements provide actionable steps or solutions based on the normative goals, guiding what should be done to improve or change the situation described in the empirical findings. Essentially, empirical data informs normative views, which in turn lead to prescriptive actions.
How to evaluate a hypothesis before testing
Before testing a hypothesis, it’s important to evaluate its clarity, feasibility, relevance, and theoretical grounding. First, ensure the hypothesis is clear and specific, stating the expected relationship between the variables in precise terms, as vague hypotheses can lead to confusion. Next, check for testability—ensure the hypothesis can be empirically tested through observation or experimentation, using available methods and data. The hypothesis should also be relevant to the central research question, aligning with the study’s objectives. Additionally, it must be grounded in existing theory or prior research to explain why the relationship between variables is expected, guiding how results are interpreted. Finally, ensure the operational definitions of the variables are clear and measurable, and that the research is feasible in terms of data availability, resources, and time. Evaluating these aspects will help determine if the hypothesis is ready for testing.
Likert Scale
Likert scales are useful because they allow researchers to capture varying levels of sentiment on a topic, enabling more nuanced insights into respondents’ attitudes. They are widely used in social sciences, marketing research, and political studies. The results can be analyzed to quantify opinions or measure the intensity of agreement/disagreement on various issues.
Phraseology:
The way questions are worded (phraseology) can significantly affect responses. Ambiguous, complex, or leading questions can confuse respondents or influence their answers. It’s important to use clear, concise, and neutral language that does not sway the respondent in any particular direction. For example, a question like “Don’t you agree that global warming is a serious issue?” is leading and can bias responses. Instead, rephrase it as “Do you think global warming is a serious issue?”
Question Order:
The order in which questions are presented can also affect the responses, a phenomenon known as order bias. Respondents may be influenced by the questions asked earlier, leading them to answer subsequent questions in a particular way. For example, asking about political views before asking about specific policies might prime respondents’ opinions on those policies. Randomizing question order or grouping related questions can help reduce this bias.
Filter Questions:
Filter questions are used to determine whether a respondent should answer a particular set of questions based on their previous answers. Poorly designed filter questions can confuse respondents or lead to inaccurate data. For example, asking someone who does not drink alcohol about their preferences for different types of alcohol may not be relevant, but without a proper filter, this question could be included unnecessarily.
Open-Ended vs. Close-Ended Questions:
Open-ended questions allow respondents to answer freely, providing more in-depth information and capturing nuanced responses. However, they can be difficult to analyze quantitatively and may be time-consuming for both the respondent and the researcher. For example, “What do you think is the most important issue facing society today?” allows for a wide range of responses.
Close-ended questions offer predefined answer options (e.g., yes/no, multiple-choice, Likert scales) and are easier to analyze statistically. However, they may limit the richness of responses. For example, “Do you think social media is harmful? (Yes/No)” restricts the respondent to a binary choice.
Non-Responsive Issues:
Non-response refers to situations where respondents fail to answer certain questions or the entire survey. Non-responses can be problematic because they may lead to bias if certain groups of people are more likely to skip questions. This can result in incomplete data and skewed results. To address this, it’s important to ensure that questions are engaging and relevant, and to consider offering incentives or reminders to encourage full participation.
Population
refers to the entire group of individuals, items, or observations that a researcher is interested in studying or drawing conclusions about. This group is typically defined by specific characteristics relevant to the research question.
Event Data
Event data refers to information collected based on specific occurrences or events. This type of data focuses on recording or tracking occurrences, often used in studies of patterns, trends, or outcomes tied to particular events. For example, tracking voter turnout on election day or the impact of a specific policy change.
Textual Data:
Textual data involves data collected in the form of written words. This can include responses from open-ended survey questions, social media posts, or documents. Textual data is often analyzed through techniques like content analysis, sentiment analysis, or natural language processing to identify patterns, themes, or meanings.
Experimental Data – Placebos, Paired Testing:
Placebos: In experimental research, a placebo is a substance or condition that has no therapeutic effect, used as a control to test the effectiveness of a treatment. For example, in drug trials, one group may receive the actual drug, while another receives a placebo, allowing researchers to assess the drug’s effectiveness.
Paired Testing: Paired testing involves comparing two related conditions or groups that have been matched or paired in some way. For instance, comparing the performance of the same group of participants before and after an intervention, allowing researchers to observe the impact of that intervention.
Non-Obtrusive Measures:
Non-obtrusive measures refer to methods of data collection that do not interfere with or alter the behavior of the subjects being studied. These methods are often used in natural settings where researchers aim to observe subjects without affecting their actions. Examples include observing consumer behavior in a store or analyzing social media data without interacting with users.
Focus Groups:
Focus groups involve a small, diverse group of participants who are interviewed together, often guided by a facilitator, to discuss a specific topic. This qualitative method allows for in-depth discussion and exploration of ideas, opinions, and perceptions. Focus groups are useful for understanding attitudes, behaviors, and perceptions in detail.
Multi-Mode, Triangulation:
Multi-Mode: Multi-mode refers to the use of different methods or channels for data collection within the same study. For example, combining surveys with interviews or online questionnaires with in-person focus groups to gather a range of perspectives and data.
Triangulation: Triangulation is a method used to increase the credibility and validity of research findings by using multiple data sources, methods, or theories. This approach helps to confirm or cross-check results from different angles, reducing the likelihood of bias or error.
Split Ballot Experiments
Test how different question formulations or conditions affect survey responses by providing different versions to different groups. They help identify which phrasing or format yields more reliable results.
Pre-tests
are trial runs of surveys or experiments conducted with a small group to identify issues like unclear questions, technical problems, or timing concerns before the main study.
