mean and standard deviation only work well for ______ ________ ____.
Normally distributed data
- data can become skewed - mean and standard dev. don’t work as well
Why can skewness be problematic?
- can pull the mean away
- not robust to extreme values
- data can be skewed left or right
What is modality? How is it challenging?
mode = value that appears most often
- shows peak
- can have more than one peak (uni/bi/multimodal)
- challenging for researchers - what is the typical value?
Why are research methods necessary?
Good research design is incredibly important, for example the prefrontal lobotomies that turned people into vegetables for no goof reason
What is descriptive research? (under observational research)
- describes aspects of a phenomenon, population, etc.
- tries not to interefere with how it’s arisen
- just cataloguing things
What is correlational research? (under observational research)
- looks for statistical relationships between variables, phenomena, etc., make generalizations
- finding assocaiations between varaibles
- often uses methods of descriptive research to gather the data
What is experimental research
- attempts to manipulate variables to find casual roles between them
- critical difference between this and observational research (causes, other than just predicts)
What is naturalistic observation?
The recording of behaviour without trying to manipulate it
- eg. warfare is not distinctly human
- limited - can’t make casual statements, but can clue you into new hypothesis
- obseravtions in real world; high external validity (findings are valid in real-world settings)
What is reactivity?
- observations can effect behaviour
- if you wait long enough, will subside naturally
- some methods can help disguise, reduce reactivity caused by prescence
Describe case studies
- weakest of all research designs
- examines one or a few people in great detail, typically over a long period of time
- useful for existence proofs (demonstration that a given phenomenon can occur)
- doesn’t control variables, so it’s not ctually “proof”
- can be highly subjective
- useful for very rare or unusual phenomenon difficult to study in the laboratory
What is the fundamental flaw with anecdotes?
- short, personal experience
- used as evidence in arguments
- does not establish causation
- no control over other possible reasons, not necessairily able to be generalized - survival bias
- may not be representative
- may be ignoring contradictory claims
- *are not evidence
Describe the research methods of surveys and self-reports. What are some of its fundamental flaws?
- questionnaires, interviews
- easy to collect large amounts of data on numerous factors
- does not establish causation between variables
- doesn’t allow much explanation as to why - can hint at possible causes
- can be massively influenced by wording
- there is an assumption that the person understands the question being asked
- assumption that the person has insight and are being honest
- positive impression management - trying to make themselves look better
- malingering - trying to appear disturbed in a way
- lapses in memory/biases can affect responses
Describe random sampling.
- important to consider how generalizable the data is
- ensures every person in a population has an equal chance of being chosen to participate
- need to consider the method in which it is administered
- eg. email doesn’t include as many senior citizens
- required for generalizable results in any form of research
What is reliability in scientific testing?
- consistency of measurement
- test re-test reliability
- if you re-test, should produce same or similar results - shoudl be able to replicate
- interobserver reality
- two people taking the same test should arrive at the same conclusion
- also called “interrater reliability”
What is validity in scientific testing?
- the extent to which a system measures what it says it does
- eg. lie detectors don’t detect lies, detect physiological responses
Describe correlational designs
- used to examine the relationship between variables
- relies on the “correlation coefficient”
- a measure to which the degree in which two variables are related
- closer to 1/-1, more drastic
- denoted as p (greek rho) or “r”
- positive - increase in x and y
- negative - decrease in x and y
- 0 - not relationship
- unless you have a perfect correlation, there will always be an exception
Why can’t we do research methods intuitively?
- people are notoriously bad at estimation
- illusionary correlation - perception of correlatioon where there is none
- confirmation bias - seek out only infomration that supports your expectations
- availability heuristic - estimate likelihood depending on how easily it comes to mind
- correlation does not equal causation
- the third variable problem - two variables may be related to one another because they are both casually related to a third variable
How do you find causation?
- observational - too many factors, cannot determine causation
- experimental - less variable factors - results are more likely to be due to the manipulated variable
- means are useless without the variation around them
- can use error bars
confidence interval - why it’s important to replicate studies
Why does data fluctuate?
- impossible to control all aspects of an experiment (either practically or ethically)
- generates unsystematic variation (aka error) in the measurement
- error bars can represent different things
What is statistical significance?
- probability of obtaining the result is less than a pre-designed threshold
- does not mean it’s practically significant
- with a large enough sample size, all findings are statistically significant
List some hazards in experimental design
- random assignment is not used (randomly sorting subjects into the experiment’s groups)
- assumes that uncontrolled differences are evenly distributed
- don’t confuse with random sampling
