Review of essential theory Flashcards
what choice of statistic?
- scale of measurement
- research aims
- experimental design
- properties of dependent/outcome variable
scales of measurement
- nominal
- ordinal
- interval
- ratio
categorical
nominal
discrete or continuous data
- ordinal
- interval
- ratio
nominal
- numbers or names serve as labels e.g. gender/religion (numbers = allocating numbers for a category)
ordinal
- data organised by ranks
- values represent true numerical relationship
- intervals between values may not be equal
e.g. race position, likert scale
interval
- true numerical relationships and intervals between value are equal
- no true zero
- e.g. temperature, shoe size
ratio
- true numerical relationships
- true zero
- most accurate
- height, distance
research aim: decribe
- summarize a set of sample values
- typically use just two stats : central tendency, spread
- e.g. average, spread, shape
- discrete or continuous data
- normally distributed
- use mean as measure of CT
- use standard dveiation as measure of spread
discrete or continuous data not normally distributed
- use range as a measure of spread
- use median as measure of CT
categorical data
- measure of CT: mode
research aim: infer relationships
- relational research explores relationship between observed behaviors or phenomena nothing is actively manipulated
- can’t infer causality but can describe relationships
research aim: infer differences
- experimental research (influence of IV on DV)
- can make claims about causality IF we control confounding variables e.g. counterbalance, random allocation
Independent variable
- hypothesized to influence the DV
- known as factors
e.g. drug, age group - always measured on a categorical scale
dependent variable
- outcome variable
- hypothesized to be dependent on IV
- e.g. test,scores, reaction time
- measured as discrete or continuous
levels of IV’s
- at least 2 levels
TYPES OF LEVELS
- true-experimental actively manipulated e.g. random allocation possible
- quasi-experimental where IV reflects fixed characteristics e.g. right or left handed
between subjects design
- independent groups
- each participant only in one level
within subjects
- repeated measures
- all subjects take part in all levels
mixed designs
- at least 1 IV is between subjects, at least one IV is within
e.g. handedness (between) and game (within)
experimental designs: test of differences: 1 IV - 2 levels
- between Ps: independent t-test
- within Ps: paired t-test
experimental designs: test of differences: 1 IV > 2 levels
- between Ps: 1-way independent ANOVA
- within Ps: 1-way repeated measured ANOVA
experimental designs: test of differences: 2 IVs (factorial designs)
- between Ps: 2-way independent ANOVA
- within Ps: 2-way repeated measures ANOVA
- mixed: 2 way-mixed ANOVA
properties of normally distributed data
- symmetrical about the mean / no skew
- bell shaped
- mesokurtic
mesokurtic
- positive kurtosis value
- sharper peak
platykurtic
- negative kurtosis value
- extreme values considered not normally distributed / parametric tests not appropriate
leptokurtic
- small s.d.
- extreme variation cannot be considered normally distributed/ parametric tests not appropriate
- positive kurtosis value
positive skew
falls towards the more positive value
bimodal data
2 modes in the data
uniform data
all data equal
what are statistics for
to draw inference, say something about a population
z-scores
- scores from a normally distributed population
- 95% of values lie within +- 1.96 s.d. of mean
mean of sampling distribution of the mean
equivalent of population mean
SDM (sampling distribution of the mean)
plot of all possible sample means
- normally distributed
- SDM mean equivalent to population mean
- standard error = s.d. of SDM
SE (standard error)
s.d. of sampling distribution of the mean
what happens to SE as sample size increases
SE decreases
ESE (estimated standard error)
an estimate of the standard error, based on our sample
Confidence Intervals
interval estimates of population parameters
- typically 95% CIs
Calculating confidence intervals for sampling distribution of the mean (when you don’t know populatoin s.d.)(around a sample mean)
- critical value of t where 2.5% of scores are higher/lower
null hypothesis (H0)
there is no difference between population means
- always assume is true
p-value
the probability of measuring a difference of that magnitude if the null hypothesis is true
- between 0-1
alpha (a)
threshold of probability where we will be willing to reject null hypothesis
~0.05
if p>a reject null
result not from chance alone
Tyep I error a
- rejected null
- null hypothesis true
Type II error β
- fail to reject null hypothesis
- null hypothesis false
