We'll spend the end of the class peer-explaining our proposals bc we will have two weeks prior to our next class and this will help us move along in the interim.
What's "real"
has to do with whether there is a natural world
observable phenomena
self-referential
measurable
Kuhn-->HOW to build knowledge
Popper -- knowledge has to be falsifiable
since this is a class in human subject research, we are seeking things that are measurable
can we observe it in some way that has some truth value to the underlying phenomena<--this is where the validity comes in
MT: this relates to the abstract I did on Borgman's article for Leah's class
reification
an experiment/interview, etc. produces observations
correlation is good, but what we really want to do is explain things
we can do this in two general ways: idiographic, nomothetic
she wants us to do nomothetic work to understand how you can get at broader explanations
try to ask questions in ways that you can get a broader view of what might be going on
causality
1. cause precedes the effect
2. cause is correlated to the effect
3. no alternative explanation for the effect (non-spurious)
the bigger the amount of data doesn't mean sthg is harder to find (security)
spurious correlations-->law of large numbers and correlation-->coincidence, not cause
always trying to control variance to isolate
Different notions of causation
Reliability
Validity
internal
external
This is John Stuart Mill:
Grounded theory-->L. Starr is the best in our field on grounded theory
say we have 70 interviews of astronomers around the world and we've asked a set of questions to all of them; we'd like to know if the difference in who releases their data have to do with different variables
we can start to map these characteristics of these people and look for order effects; we can look for waht's related bc we have enough ppl in the sample; ule out alternatives by the other kinds of questions we ask
what the grounded theory does for you is to develop hypotheses as you go; but the difficulty with any type of observation like this, you're laying out your complete picture at once (IV, DV, units of analysis, correlations we're looking for, etc.)
the risk with open-ended questions is that the patterns you begin to see, it's hard not to alter your questions, which leads to the introduction of a lot of variance, risking validity and reliability
grounded theory says you iteratively test again to ensure the patterns exist throughout
once you're off the path, it's indefensible
Necessary and sufficient model
often used as an explanation for what it takes for sthg to be causal, but it's not as powerful a set as the causality requirements
for a to cause b, a must be present for b to occur and a must be sufficient for b to occur
philologist working bt Chinese Buddhist texts
the notion of translation: literal/faithful, or recasting sometimes as a cultural transmission or a sacred translation; the notion of translator was different, and need to know the context to understand the authority by which these translations are thought of as equivalent
EPSM: equal probability of selection method
drawing conclusions abt individuals based on observations of groups-->ecological fallacy
Bias
there is a whole lot of provenance involved with reliability-->want to explain why aspects are done in the specific way you've designed so ppl can replicate with sensitivity to the same concerns-->has implications for repurposing and combining data in the future
Validity
Calvin Trillin article
what are the factors that determine whether ppl adopt FB and use it more than an hour a day
sample of 10k ppl and cover all age ranges, demographic groups, etc.
will that have higher external or internal validity
internal validity: how well does the measure reflect the manipulation within your study within the population you've tested
tightly controlled variance, but more restricted population
external validity: how well the measure and the protocol will translate to other contexts and other populations
less controlled variance, but broader population
reliability resides in the operationalization
cleanup-->look for spurious correlations, ppl who were messing with you, attrition, etc.
Research design
Next time we'll get into conceptualization, operationalization, measurement; categorical, ordinal, interval, ratio; units of analysis
Topic | Day | Article | Workshop |
3 | 1/27 | Diana | |
4 | 2/10 | Mario; Morten | Diana; Kathy |
5 | 2/24 | Dustin | Natasha; Mario |
6 | 3/3 | Rob | Dustin; Marika |
7 | 3/7 | Natasha | Morten |
8 | 3/10 | Kathy; Marika | Rob |
9 | 3/17 | Milena |