Chapter 8 Bivariate Correlational Research Flashcards
3 goals of the scientific approach!!!
describe
predict
explain
3 big types of research designs
descriptive, correlational, experimental
examples of descriptive studies
surveys, nat. obs., case studies
examples of correlational studies
questionnaires, interviews, observational measures
each research design is associated w/ which scientific goal?
descriptive/describe
correlational/predict
experimental/explain
bivariate correlation
association involving exactly 2 variables
main difference b/w quantitative and categorical variables
categorical numerical values are arbitrary, and quantitative numerical values are somewhat ordered (less to more, lower to higher, etc)
2 types of quantitative variables
discrete: no decimals (ex: # of books)
continuous: unlimited number of values b/w adjacent values (ex: reaction time of 1.37 seconds)
3 types of correlational coefficients and when they’re used
Pearson’s r: 2 variables at ratio/interval level
Spearman’s rank-order r: 2 variables at the ordinal level
Point-biserial r: 1 variable w/ 2 categories and 1 continuous variable
what tools are useful when the IV is categorical and the DV is quantitative?
bar graphs, mean, t-test
effect size
magnitude or strength of a relationship b/w 2 or more variables
which effect size is usually more important, large or small?
large. more accurate predictions
R-squared
proportion of variance shared by 2(+) variables
what does a narrower CI indicate?
the more precise the point estimate may be
sample size in terms of stability
a larger sample size gives a more stable estimate of effect size than a small sample size
p-value
likelihood that the association is due/not due to chance
p < 0.5
significant, unlikely that the association is due to chance
p = 0.5 or p > 0.5
not significant, likely due to chance
is a p value a correlation coefficient?
no
when the 95% CI does not include zero
p<0.5 and the correlation in the sample is unlikely to have come from a population where the corr. is 0
when the 95% CI includes zero
p>0.5 but we still can’t rule out that the true population corr is zero
what effect size is more likely to be statistically significant?
larger/moderate
a very small effect size might be statistically significant for what?
large sample
what does replication test?
consistency
define outlier
a score/point on the graph that is highly deviant from the rest of the data
online vs offline outliers effect on correlation
online: inflate coefficient
offline: reduce coefficient
which samples are the most affected by outliers?
small samples
2 ways to detect outliers
- 3 SDs away from the mean
- median absolute deviation
2 ways to handle outliers
- remove from dataset before inferential statistical analysis
- keep in dataset and recode w values equal to that of 3SDs from the mean
restricted range
when the sample under the study doesn’t include the full range of variables
what does it mean if a sample is homogenous?
the values of the sample are all pretty similar (creates a restricted range)
curvilinear association
the relationship b/w 2 variables is not a straight line and r=0. U shaped curve
3 criteria for causation
covariance
temporal precedence
internal validity
the 3rd variable must be associated with what to be considered a potential alternative explanation?
both variables
spurious association
the apparent correlation b/w X and Y is actually caused by Z
moderation can address which validity?
external
moderation
the strength/direction of an association b/w A and B differs depending on the level of C (moderator)
