choosing a statistical test Flashcards
what are 3 factors to choosing the right statistical test
-if ur looking for difference or correlatiom
-if its a difference, what experimental design
-the level of measurement
why are statistical tests important
-determine if a signficant difference or correlation exists
-and consequently wether the new hypothesis should be rejected or retained
how do yk if its a correlation or difference
-check wording of hypothesis
which designs are related
-repeated and matched pairs
what designs are unrelated
-independant designs
what is quantitative data level of measurement
-can be divided into diff levels of measurement
what is nominal data
-categorical
-data only appear in one category(discrete)
-theres no order
-e.g young vs old
their fav fb team
tall vs small
what is ordinal data
-data ordered and ranked in some way
-the difference between them is not meaningful or fixed
-subjective
e.g asking someone to rate psych on scale of 1 to 10
what is interval data
-data measured on a scale
-unit of measurement
-precisley defined
-temp, length, time
-better than ordinal bc more detail preserved as scores not converted to ranks
what are parametric tests?
-c related t test and unrelated t test, pearsons r
-more powerful and robust
-detect significance within some data sets
what are the 3 criteria parametric tests must include
1.interval - use actual scores rather than ranked
2. must draw data from a normal distribution for the variable being measured (not variables w a skewed distribution)
3.homogeneity of variance - sets of scores in each conditon should have similar spread
(one way to determine is comapring SD in each condition)
if similar parametric test can be used
(in related designs its assumed the 2 grps have similar spread.
look at the diagrams in revision folder
