Quasi-Experiments, Small N Designs and Big Data (Week 13) Flashcards
What are the major differences between experimental and correlational research?
Exp res includes at least one manipulated variable
Exp res usually involved random assignment to different levels of a manipulated variable
What is a quasi-experimental design?
Provide examples
Involves an IV and a DV but participants are not randomly assigned to levels of an IV - there is no manipulation of the IV
Ex: studies on gender differences - gender is not randomly assigned
What are selection effects? - give an example 2 explain
What validity does this threaten?
How could you combat this effect?
When testing the success of traditional vs. nontraditional schools it is possible that higher achieving families send their children to nontraditional schools explaining effects
Internal validity
Matched groups - ensure that average SES is the same across both groups
What is history threat?
External, historical event that happens for everyone in a study at the same time as the treatment variable
ex: Autism prevalence rates increase coincides with increases in child vaccination
What is a demand characteristic?
Why is this often a problem with quasi-experiments?
Participants who know they are in a treatment group expect to get better
We cannot use a blinded design
ex: participants selected to undergo surgery and know they are in treatment group
What is a small-N study?
N usually refers to the number of participants - small-N means a small number of people
Case studies involve only one participant studied in great detail
What are some examples of famous case studies discussed in class?
- HM - Henry Molaison, unable to form long-term memories
2. Genie - feral child, raised in complete isolation until middle childhood
In evaluating treatments with small-N, what is a stable baseline? what is a reversal design?
Stable baseline - having a long waiting period of assessment before introducing the treatment
Reversal design - after providing treatment, the treatment is removed to see if performance decreases
Define qualitative research? Name the different types
Hint: there are 6
Research that identifies themes instead of using statistics
Types: one-on-one interview, case study research, focus groups, record keeping, ethnographic research, process of observation
What are the pros and cons of small-n studies?
Pros: high detail on a small number of participants, able to draw conclusions about a very specific population, some research topics are challenging to study and require small-n
Cons: lack of generalizability, much harder than statistics based, tends to be ignored by most mainstream psych journals
What are big data studies?
Collecting and analyzing data on thousands or millions of individuals
Data is usually collected online
What is machine learning?
Provide an example
Using algorithms and massive computing power to identify patterns in big data
- usually involve training an algorithm to perform a task with accuracy
ex: suicide detection based on social media
Are machines unbiased?
Real world example?
No, bias occurs in machine learning just as it does in traditional experiments
Ex: amazon recruiting tool biased against women - resumes that mentioned the word women were heavily penalized, also facial recognition works best on white men
What is an ethical concerns associated with big-N studies?
Web-scraping - the process of extracting online data usually through social media posts
What are the pros and cons of big data and machine learning?
Pros: generalizability and statistical power, easy to access billions of posts, can detect subtle trends (such as attitudes during an election)
Cons: bias may be difficult to detect, so-called “black box” effect: the solution is so complicated humans can’t understand it, privacy concerns (whether people have properly consented)