Q10 Ch 30 Comparative Analysis

comparative stats

tests that compare characteristics of 2+ independent populations OR compare before/after characteristics of the same pop being followed over time

1st step + key analysis: case-control

1st step: cases+controls are similar EXCEPT disease status

Key Analysis: odd ratios → see if case/controls have diff exp. histories

1st step + key analysis: cohort

1st step: exposed+unexposed are similar EXCEPT for exposure status

Key Analysis: rate ratios → exp/unexp have diff rates of incident disease

1st step + key analysis: experimental

1st step: individuals on an intervention+control groups are similar EXCEPT exposure status

Key Analysis: tests of efficacy → intervention/control groups have diff outcomes

null hypothesis / H₀

expected result of stat test - if there’s NO diff betw. 2+ values being compared

null result / H_a

no statistically significant difference

alternative hypothesis

expected results if there IS a diff betw. the 2+ pop being compared

reject the null hypothesis?

values ARE DIFFERENT

reject the idea that values have no difference

*thinking there is a difference when the null is true = type 1 error

fail to reject the null hypothesis?

NO evidence that values are different

*saying there is no difference when there IS difference = type 2 error

p-value

aka probability value - likelihod that a test statistic is as extreme/more extreme than the one observed would occur by chance if null was true

small p-value → observed test is likely to occur by chance = exposure is likely to cause a difference

type 1 error affects our result interpretation? how to reduce probability of type 1 error?

REJECTING A TRUE NULL! thinking there is a difference when there is not! false positive

increase sample size, lower alpha (stricter for errors)

significance level

p-value where null is rejected

statistical significance

p-value is less than significance level

standard significance lvl most commonly used in studies

0.05/5%

assumption, what do some stat tests assume?

assumed to be true

6 steps for hypothesis testing

1) select variables to compare

2) specify goal of test

3) check variable types

4) choose appropriate test for variables

5) confimr assumptions of tests are met

6) run test + interpret results

parametric test

variables being examined have particular distributions

nonparametric test

does NOT make assumptions ab distributions of responses

typical variables + distributions of variables in para vs. nonpara

para: used for ratio + interval variables w norm dist.

non-para: used for ranked variables + when dist. of ratio + interval is not normal

chi square test - test formula, sample null/alt. for a chi square - independence, proper reporting

• compares proportion of responses to a nominal vari to a selected value

• X² test

• tbh idk

purpose of t test

to find significant difference betw. 2 sets of data

one sample t test

compares mean value of a ratio/interval vari to a selected value

ex: average water consumption in LV → national av. is 3 liters

independent populations

pops w no individual is a member of more than 1 of the groups being compared

independent samples t test

aka 2 sample t test - compares mean values of ratio/interval variable in 2 independent pop

test used when stats are being evaluated:

R/I: mean

Ordinal: median

tests when stats in one pop is diff from a hypothetical value

one sample t test

test used when stat differs in 2 pops

independent t tests

test used when stat differs in 3+ pop

one way ANOVA

paired data

vari linked together for analysis- if theyre simialr

matched pairs t test

values of ratio/interval vari in members of 1 pop measured twice (before/after)

ANOVA

analysis of variance

compares mean values of a continuous variable across independent pops

one-way ANOVA

mean values of ONE ratio/interval predictors across independent groups of ppl

ex: mean age of pts at 3 diff family practices

uses F test

two - way ANOVA

aka factorial ANOVA - mean values of R/I vari across rgoups defined by 2 diff vari

e: mean ages by sex + smoking

repeated-measures ANOVA

compares values of R/I vari across several points i time/ several individ. matched pop