Final Deck 2 Flashcards
Define and contrast descriptive vs. inferential vs. analytical statistics
- Descriptive stats summarize/organize/simplify data from a single sample or population (ex: using mean, median, mode, variance, SD)- summarize data simply
- Inferential Stats- lets us compare two or more samples to make generalizations about the populations that were pulled from (t-test, anova, correlation)-draw conclusions based on observations
- Analytical stats-goal is to gain insight into the processes that drive the observed data patterns (ex: cluster analysis, regression analysis etc.
What is the difference between a population and sample?
- The population is the entire set of individuals of interest in the study (all caregivers)
- The sample is the set of people selected to represent the population (the caregivers selected to participate in the study)
Simple random sample and it’s primary advantage
- Everyone in the population has an equal chance of being selected for the study
- the advantage is that this is the least biased because it is intended to look like the normal distribution
- This increases the validity (accuracy) of the study because it is representative of the population
What is the difference between probability and non-probability samples?
- A probability sample is when each member of the population has a chance to be selected
- A nonprobability sample is when sampling is not completely random-easier to do, but a chance for systemic bias to creep in
3 examples of probability sampling
- Simple random: everyone in the population has an equal chance of being selected
- Cluster: selecting people who tend to congregate in specific areas (ex: going to DDI for a study about Autism)
- Stratified: tailoring your sample to be proportionate to the population you are targeting (ex: if there are 2 boys to 1 girl diagnosed with ASD, then you should be copying this proportion in your sample)
- Systematic: every “nth”. ex; every 5th person that attends speech this week
- oversampling: you might oversample to look more into a specific quality
3 types of nonproability sampling
- Convenience: You ask whoever happens to be around you to participate
- Purposeful: You judge someone to seem like a good source of info, then ask them (ex: parents on the PTA- you choose this bc you think they have special knowledge)
- Snowball: You send it to 3 people, ask them to send it to 3 more, etc.
- Self selection: You just put it out there and see what happens
Why is sample size important for a study in terms of sampling error/probability?
- A larger sample size is important for a study because it reduces the chance of bias.
- The more people you have in your study, the more likely the sample is to look like the population.
- Also, the more people in the study, the less chance of outliers coming in to skew your data. —> this is the law of large numbers
Rule of thumb on minimal sample size
- Ideally you want 300+ participants for a survey, but 100 is the minimum
10% rule of thumb-researchers should not change the sample size more than 10% than the predicted SPSS value
How about sampling for qualitative designs?
- You can have a smaller sample size for a qualitative design because the study is more reliant on the researcher’s judgement than statistical analysis.
- You’re not really trying to generalize the population in this case
What is saturation in relation to qualitative sampling?
- Saturation: When additional responses no longer add variability to the results → over time, they get no more new information from the data (get the same thing over and over again)
- people keep saying the same thing and there is no more variation
List/describe 3 features specific to qualitative research designs
- Descriptive: Since numeric data alone are insufficient, social behaviors are described in detail and understood in the context of design and usage
- Social: Focused on social phenomena
- Systematic: Data collection is procedural and consistent
- Authentic: Data collection takes place in contextualized, naturalistic settings (i.e., not in a laboratory or controlled setting)
Define observations (consider direct/indirect and structured/unstructured)
- Examination of an object, process, or other phenomenon for the purpose of collecting data about it or drawing conclusions
- is the observation direct or indirect (social media, transcript, narrative, etc.)
- Is the observation naturalistic (in the participants natural setting?) or structured (in a lab setting)
What is a participant Observation- how is it unique?
- A participant observation is when the researcher is also immersed in the activity and experiencing the activities and interactions that participants do as well.
Define interviews
- A series of questions designed to elicit information a researcher is interested in.
- Interviews have 3 types: structured, unstructured, and semi-structured
Self-Disclosure in an interview
- Interviewers frequently engage in the process of self-disclosure, where they review their collected data in light of their own background to prevent bias.
Define focus groups and contrast with interviews
- A focus group is about 6-12 people that engage in a discussion lead by a moderator.
- Focus groups highlight narratives (not behaviors).
- They use a phenomenological approach (focusing on the participant’s perception of a construct)
- interviews follow more specific questions and answers where a focus group is more of a facilitated group discussion
What is a qualitative tradition? and vs. techniques
- Traditions are much broader → they are collections of ideas, theories and data collection techniques.
- A single tradition can have multiple data techniques (case study, ethnography, grounded theory approach etc → these things can use multiple techniques to collect their data)
- The technique is the specific way you collect data
Biographical Study and one feature
- An in depth investigation of one individual over time.
- Characteristic: researcher asks exploratory/open ended questions that allow the participant to share their thoughts and build a narrative
Case Study and one feature
- A case study focuses on a specific “object” (person, topic, location, or event). - Data is gathered from one or a small number of cases, and is usually more structured than a biographical study, with more scripted and researcher-directed questions
- note that one participant is studied at a time, but there can be multiple participants in the study.
What is ethnography and one feature
- Uses interviews and observations to explore a cultural/social aspect
- it focuses on a cultural or social stuff rather than a single phenomenon
- *It does not require a formal research question or hypothesis
Grounded Theory
- This design is meant to be a more systematic qualitative tradition “grounded” in data to maximize researcher’s theoretical sensitivity, meaning, their focus on data and hypotheses testing.
- Data is coded to make comparisons across sets
3 Coding Themes of Grounded Theory
- Open Coding: Break all of the transcript data into themes (sometimes referred to as “codes”), typically by identifying common keywords in transcript sections
- Axial Coding: Identify relationships between themes/codes (causal, temporal, spatial), and group into Categories
- Selective/Core Coding: Use these relationships to identify a small number (usually one) overarching Core Category/Categories
Mixed matched surveys
- MMR combines at least one quantitative and qualitative technique (strand) in the same study.
- For example, a MMR study might employ both qualitative interviews and quantitative surveys.
Descriptive vs. Inferential/Analytic Research Methods
- Descriptive- describing variables
- Inferential/Analytic- comparing variables
Case Report
- They are detailed, retrospective description of a disease or a medical condition
- Usually just a description of one person
Case Series
- Multiple case reports on individuals with the same disease or condition- multiple people involved
Describe cross sectional studies and pro/cons
- A descriptive “snapshot” of variables for a singular measurement in one instance in time- it measures the distribution of variables only
- pros- quick, inexpensive, avoids participant attrition/mortality
- cons- can examine prevalence only (not incidence), so you cannot establish a causal relationship, vulnerable to recall bias, misinterpretation
Prevalence vs. Incidence
- prevalence- how many people have something at one point in time
- incidence- how many people are expected to have something/new cases develop
Cohort Study
- Prospective study that compares variables between two groups: one group is at risk/exposed to a particular variable, the other is controlled.
- Then you see if there is any difference in development
Case Control
- Retrospective study comparing variables between a condition vs. a control group → you collect 2 groups of people, one group has X condition, and group 2 does not.
- You go back into their history to see if there are any variables/red flags that might have something to do with the condition
Relative Risk
Associated with a cohort study
-Relative risk is the probability event in the exposed group vs/divided by the probability of event in the unexposed group
-Advantage: you can analyze incidence, gives you more info about establishing a relationship (but not causal)
-Disadvantages: more expensive, time consuming
Odds Ratio
- Associated with case control study
- It is when probability of exposure in case group divided by probability of exposure in control group
- different than with a cohort study because You are starting with who has the disorder and who doesn’t, rather than anticipating if the disorder will come up
Randomized Control Trials (RCT)
- prospective studies in which participants are randomly assigned to treatment/experimental or control groups after recruitment and assessment of eligibility, but before intervention (one intervention is used)
Crossover Designs
- A study that has multiple treatments applied to the same group → you see if there is an order effect.
- Ex: is the treatment better for people who get intervention A then B or is the treatment better for people who get intervention B than A?
Pros of using an experimental design
- good randomization reduces population bias, easier to blind/mask participants that observational studies, results can be analyzed with well known statistics, populations of participating individuals are clearly identified
Cons of using an experimental design
- $$, time consuming, volunteer biases: the population that offers to participate may not be representative of the whole population.
- Also, better at identifying cause, confounding, loss to follow-up attributed treatment
Define a type 1 error
- Claiming that a result is statistically significant when it’s not
What is p value
- The p value is the probability of making a type 1 error
- P must be less than .05 to be considered statistically significant
Alpha
- Associated with type 1 error
the estimated probability of making a type 1 error
What does a smaller p value mean in terms of type one error?
- The smaller the p value, the less chance of making a type 1 error
What is a type 2 error?
- Saying that results are not statistically significant when they actually are
Define statistical power
- Power is a function of beta, effect size, and sample size
- It is the probability that a study will correctly reject a null hypothesis (opposite of beta)
Define beta
- Estimated probability of a type 2 error set at 0.8 or 20%
- We want this probability to be less than 20%
- We want a less than 20% chance of making a type 2 error
What is statistical power used to estimate?
- The likelihood of correctly rejecting a null hypothesis
- AKA correctly identifying significant results
What is effect size?
A measure of the magnitude between variables or groups used in a study
What is the difference between a survey and a questionnaire?
- A questionnaire is just the list of questions
- A survey includes the process of surveying- it has methodology (sampling, distribution, analysis)
- Surveys include questionnaires
What is pilot/field testing?
- Pilot testing allows you to test out your administration, directions, the way questions are worded etc.
- It is important to practice it on people before the real survey is distributed
- You can still change things in your questions at this point
What is the difference between select response items and constructed response items?
- Select response items have the participant choose from previously created answer choices
- Constructed response items require the participant to create their own response
Give 3 examples of select response items and the pro/cons
- Multiple choice
- true/false
- Rankings
- Likert scales/Ordinal Scales (agree-slightly agree- disagree- slightly disagree)
- Matching
- Forced choice (one of 2 choices)
- pros: easier for participants and for us to score. Time efficient. Simple individual or group administration. Greater reliability of content and format. Simple conversions to quantitative scale
- cons: provides no background info (only qualitative can give background info)
Give 3 examples of constructed response items and the pros/cons
- Short or long answer write ins
- Writing samples or artifacts
- Oral discussion
- Short or long answer write in
- Writing samples/artifacts
- Oral discussion
- Performance
- pros: More creative, expansive responses available
- “Authentic” samples of behavior
- Better suited to assess writing skills, mechanical expertise, and other performance-based skills
- cons: more time consuming to grade
What is a leading question/how does it compromise responses?
- A leading question is a type of question that influences a particular response through the wording of the question
- Example: do you agree that voter ID laws will prevent eligible voters from casting their vote?
- it sways people to respond in a certain way
What is a double barreled question/how does it compromise responses?
- Double barreled questions using the word “AND/OR”-is very dangerous because it introduces multiple concepts
- Ex: do you like coffee and tea?
- it is a question about two things and it doesn’t create good responses
What is acquiescence/dissent bias? Name one thing that contributes to it
- tendency to agree/disagree regardless of meaning → caused by a desire to please the examiner, predilection towards/against the construct, or response options all being “keyed” towards agreement/disagreement statements, they want to just finish it quickly, they don’t understand what you are asking so they just click agree for everything
What is extreme/neutral response bias? Name one thing that contributes to this
- Tendency to choose extreme/moderate options, regardless of actual opinion
- likert scales tend to contribute to this because people want to answer strongly in either direction
- One thing you can do about this is add in a qualitative question-this allows the person to give their background as to why the answers are so extreme or realize they want to change their answers
What is social desirability? What contributes to this?
- Tendency to respond to sensitive questions in socially desirable ways (different than wanting to please the researcher- bc this is about being socially acceptable)
Can be affected by - content (construct): some traits are more socially desirable than others
- context: Knowing the respondent’s identity or consequences of results → if their identity is anonymous, they are more likely to answer honestly)
- *caused by poorly worded questions, not being anonymous and leads to an overestimate of socially desired results
What are 3 ways to control response bias?
- Manage content format (limit the number of items, use good wording, consider some qualitative items, use a combination of positive and negative keyed items)
- Manage context AKA testing situation (ensure anonymity, promote honesty, minimize distractions)
3.
What is skewness/what does it look like on a graph?
- Skewness is a lack of symmetry on one side of a graph
- caused by outliers (dragging the graph in either direction)
- It looks like the curve is shifted far to the left or right of a graph
What is kurtosis/what does it look like on a graph?
- Kurtosis is how spread out the scores are
- Looks like how wide the bell curve is