Exam 2 Flashcards
what is experimental research?
direct manipulation and control of variables
measuring a response
what is a nonexperimental research
relationships studied by making observations or measuring variables as they exist naturally
what type of design assesses the relationship among variables and does not contain a true variable?
non experimental
classification of observational designs
naturalistic observations
participant
structured- intervene to cause hard-to-observe situations and record behavior
time related sampling
time sampling- observations in only certain predetermined time intervals
event sampling-recording each event that meets a predetermined definition
situation-related sampling
situation sampling- behavior at different locations and conditions
subject sampling-individuals in certain locations
subject/participant reactivity
demand characteristics-may behave in ways the researcher wants them
social desirability bias- may act in ways they think they should act
observer biases/ errors
observer bias- researcher’s choice for which acts to observe leads to errors
expectancy bias- knowledge of hypothesis may lead them to miss certain behaviors
observer drift- criteria changes over the course of observations
to promote validity
strong observational definitions
training of observers to criterion
multiple observers
to establish reliability
codebooks to prevent drift
intterater
correlation coefficients
pearson’s R
indicates direction
weak correlation
0-0.29
moderate correlation
0.3-0.49
strong correlation
0.5-1
what does the p value do
tells if there is a significant relationship
p=0.05 we are 95% confident that there is a significant difference
report if p< 0.05
open ended questions
answer question in own words, doesn’t lead them in any direction
close-ended
forced choice
ranking scales
likert scales-how much you agree
response set
tendency to answer all questions in a certain manner
what leads to response set
social desirability
extreme response bias
response acquiescence (yea-saying”
how to avoid response set
reverse scoring items
prevent fence sitting-get rid of neutral options
what should survey questions involve
simplicity
double-barrled
loaded
negative wording
simplicity
questions avoid technical wording and jargon
double-barreled
questions should avoid asking 2 things at once
loaded
questions should avoid bias wording
negative wording
should avoid double negatives
how to avoid leading questions
make sure mutually excluisve and unambiguous
context effects
unintended influences on answers not related to the content but to the context it appears
item-order effect
the order in which items are presented may effect answers
types of non-probability sampling
convenience- those nearby
snowball- existing part. recruit others
quota- subgroups in the sample are recruited to be proportional to the subgroups in a population
self-selection- individuals take part in research on their own accord
elements of an experiment
manipulation
measurement
comparison
control
levels
categories
conditions
groups that are made up by combinations of the levels
advantages of BSDs
doing one level makes it impossible to do other levels
use with subject IVs
disadvantages of BSDs
requires lots of participants
creating equivalent groups
WSDs advantages
fewer Ps necessary
all levels consist of the same Ps
more efficient and sensitive
WSDs disadvantages
must control for order effects
yoked control groups
control group members are inked to experimental group members
control basics
only the IV should change systematically across conditions in a study
control variables
held constant
name the 12 threats to internal validity
1) instrumentation- did researchers use same standards when coding?
2) demand characteristics- were part. blind to purpose?
3) maturation- comparison group have time to adapt?
4) history- did comparison group that was also compared to internal event
5) placebo effects- did comparison group get placebo to rule out belief in drug?
6) regression to the mean- need a comparison group who was equally extreme at pretest
7) order effect-counterbalance the orders
8) selection effect- random assignment
9) observer bias- was observer blind to IV
10) testing- did levels of IV take the same test
11) attrition- did they included data only from those who took both pretest and post
12) design confound- did they control nuisance variables
between subjects
different part. at each level of the IV, tested in 1 condition
within subjects
tested in same condition
macthed groups
matched to the DV or other extraneuous variables
2 criteria for rando assign
1) equal chance of being in either group
2) assigned to a condition independent of other part.
carryover effect
an effect of being tested in 1 condition on part. behaviors’ on other conditions
types of carryover effects
practice
fatigue
context/contrast effect- being tested in 1 condition can change how part. perceive stimuli
solutions to order effects
1) counterbalancing- testign diff. part. in diff. orders
2) complete counterbalancing- equal # of part. complete each possible order of conditions
3) random counterbalancing- when # of conditions is large, order of conditions is randomly determined for each participant, WE CAN DETECT ANY CARRYOVER EFFECTS AND ELMININATE CONFOUNDS
why is random assign. important for WSDs
instead of randomly assigning conditions, they randomly assign the order of conditons
design confounds
an additional variable unintentionally varies systematically w/ the IV that then provides an alt. explanation to our result
selection effects
when our levels of our IV (condition) systematically have diff. types of ppl in them
how to control selection effects
rando assign or matched groups
or choose withing subjects (repeated measures) designs
types of BSDs
post test
pre/post test
both ensure no selection effects
post test
random assignment to IV, measure DV once
pre/post test
equivalent groups
rand assign.
get pre to esnure equivalency
apply IV to get post
track performance over time
null effects
when there is no effect of the independent variable, study finds that the IV did not make a difference in the DV (no significant covariance between two variables.)
ceiling effects
top easy
everyone does well
floor effects
too hard
everyone does bad
confounds acting in reverse
creating a null effect
masking the difference between groups b/c of confounds
DEBREIFING OR MANIPULATION CHECKS CAN HELP W/ THIS