Exam 3 Flashcards
two vs one tail
1tail: looks for increase or decrease
2tail: looks for change
4 possible outcomes of a statistical test
- ho is true and we dont reject: correct, probability=1-a
- ho is true and we reject: TypeI error, probability=a
- ho is false and we fail to reject: type II error, probability=b
- ho is false and we reject: correct, probability=power=1-b
when to perform each test
t-test= sigma is unknown
z- test= sigma is known
assumptions
- normality: each group should follow a normal distribution
- homogeneity: variances of 2 groups should be roughly equal
- independence: observations within each group must be independent and independent of each other
- scale of measurement: data should be measured on interval ratio scale, t test assumes meaningful differences between observations and an appropriate scale of measurement
F size meaning
Large F: more variance between, greater probability of rejecting null
Small F: between not large, lower probability of rejecting null
information needed to calculate independent samples t-stat
- value for each n
- value for each X^-
- value for each s or s^2
consequence of increasing sample size if all other factors are held constant
decreased standard error and increased likelihood of rejecting ho
why comparing more than 2 treatments means you should use ANOVA instead of multiple t-tests
- using several t-tests increases risk of type 1 error
- to control experiment wise alpha rate
if null is true for ANOVA
MSb should be about the same size as MSw
goal of any inferential test
explain variance in DV/ outcome variable
2 types of alpha
Alphac: comparison, conduct type 1 error in specific comparison
Alphae: experiementwise, conduct type 1 error in all/any comparison
Tukey’s HSD
method controlling experiment-wise alpha
as sample variance increases…
value of t-stat decreases
if you increase s2^2
MSw increases, size of F ratio decreases
if X^- increases
MSb decreases and F ratio decreases