week 9, validity + descriptive statistics Flashcards
how can we combine some of the elements of different non-experimental designs to make them stronger?
non-equivalents control group design with pre and post test
what is a quasi-experimental designs
interrupted time-series design
measuring a variable over time both before and after
-no control group !!
what are some threats to internal validity for interrupted time series design designs?
check slide 13 for slide 1
selection bias -> less concern bc participants comapred to themselves
potential testing effects are distributed along the entire time series
regression to the mean, one data point may be extreme but rest are unlikely to be affected
what is cross sectional design
compare language, memory, emotional response, social skills, etc
how to combat generation cohort effect when you test different ages?
longitudinal design, test same kids throughout their lives
what is time-lag design
different cohorts are tested in different years, but age is kept constant ( 12 yr olds from 3 different years)
if theres no dfference then generation nor year of test matters
what is a confounding variable for longitudinal design
year of test, could have history and testing effects
what is a cohort-sequential design, what is a risk of it?
combines cross sectional, longitudinal, time-lag and requires replication and extension.
selective attrition is still an issue
internal validity in experimental research
to what extent are changes in the DV caused unambiguously by changes in the IV
- have all alternative explanations been adequately ruled out?
what factors reduce internal validity?
- violations of random assignment (no control group, non equivalent groups)
2.presence of confounds
external validity in non-experimental research
are the results derived from a sample generalizable to whole population?
- concerned with sampling methods
external validity in experimental research
generalizability of experimental results to other people or situations (dif contexts/outside of lab)
what factors reduce external validity?
- violations of random sampling (cant generalize to pop)
-replicate findings in other samples of same pop, other pop etc - small sample sizes, over/underestimate effect
- artificiality of experimental setting (talking about political views w strangers)
laboratory observations have… internal validity and …. external validity
naturalistic obersvations have internal validity and …. external validity
- high, low
- low,high
unsure if DV changed due to my manipulation of the IV or not (due to low internal validity)
but !! im pretty sure it would generalize (due to high external validity)
how do we increase generalizablity of results? what is this called
combine internally valid experiments with externally valid field studies, method triangualtion
generalizability of results is not always the goal, externally invalid but internally valid studies can answer important theoretical questions that can affect human behaviour inside and outside the lab
on to slide 2
what is descriptive statistics and inferential statistics?
- describes and summarizes data
- draws inferences from data
what does a histogram plot?
frequency of occurrence
what is the central tendency?
a measure of what value the individual scores tend to center on.
tells us about the whole sample
what are types of central tendency?
- mode -> most common value (used for categorical variables, nominal scales), not useful with continuous scales
- median -> middle value, useful for ordinal scales, interval, and ratio but NOT nominal
3.mean -> arithmetic average,
mean median and mode can have the same values but normally don’t
what does it mean for a distribution to be right skewed?
- some values are extremely high
what does it mean for a distribution to be left skewed?
- some values are extremely low
use the stroop test to define congruous condition
Color word matches the ink color, easy with fast response
use the stroop test to define incongruous condition
color word doesn’t match the ink color, harder with slower response
means are affected by the distribution, therefore it is sometimes better to use the median
what is variability in the data?
- how clustered are the data?
2.how different are the scores?
what are measures of variability?
- range -> between highest and lowest value
2.variance -> average squared deviation from the mean, average squared distances -> variance.
we only care about magnitude of variance
3.standard deviation -> average deviation from the mean
we only care about magnitude of variance
square roots variance to restore metric to og unit
what is the relationship between variables in a scatterplot
the 2 variables move together, both up or both down
or
they move opposite direction,
one up one down
what does correlation coefficient (r) care about
- relationship between x and y
- size -> strength of relationship
-sign -> direction of relationship
if r=0 then what kind of relationship?
zero relationship
-1 ≤ r ≤ +1
-1 = negative perfect
+1 = positive perfect
what does it mean for 2 variables to be related?
tells you there is a relationship between variables (but not why)
one possibility is a casual relationship, give an example
paying nurses more causes an increase in household incomes
finding a correlation does not imply that one variable causes the other, what does this mean?
correlation does not equal causation
what are difficulties interpreting correlations?
- third-variable problem: another variable caused the change
2.Directionality problem: dont know if variable x caused variable y or other way
- Selection bias:systematic differences between the comparison groups
what are spurious correlations?
third variable problem & selection bias often lead to correlations that do not indicate a casual relationship.
what is appropriate language to describe a correlation?
- association
-relationship
-correspondence
-correlation
what is inappropriate language to describe a correlation?
-determine
-cause
-lead
-shape
-influence
we conduct experiments to say something about causation
