Chi Square Flashcards
what is a commonly used observational study?
exposure-disease studies
what does the contingency table record?
the frequency of each of the categories
if there are 2 categorical variables, what test are we running?
Chi square
if there are 2 continuous variables, what test are we running?
correlation/regression
if there is one categorical and one continuous variable, what tests can we run?
t test, WSR, MWU, ANOVA, KW, Friedman
what variable goes into the columns?
the disease (DV)
what variable goes into the rows?
the exposure (IV)
when data on 2 categorical characteristics are collected from one randomly selected sample, what test is run?
chi square
what is the key ? of interest with chi square test of independence?
the independence of 2 characteristics
t/f: contingency tables focus on the dependency/association bw column and row variables
true
what is another name for the test of independence?
test of association
a PT collected data from 200 randomly selected pts. pts were asked if they had a TKA in one leg within the last 10 years and if they had contra knee pain. she wants to determine if the TKA and contra knee pain are independent.
what test should be run?
what is the H0 and Ha?
chi square
H0: TKA and contra knee pain are independent/not associated
Ha: TKA and contra knee pain are associated/not independent
what are the hypothesis for chi square?
H0: 2 categorical characteristics are independent (no association)
Ha: 2 categorical characteristics are not independent (association)
are the hypotheses for chi square one or two sided?
two sided
how do we find X2? (don’t really need to know)
get the expected frequencies for each cell
calculate the o-e^2/e for each cell
add up the above values
if X2=6.048, df=1, p=0.014, what is the conclusion of the TKA example?
a chi square test of independence was performed to examine the association bw TKA and knee pain. results indicated that the association bw these variables was significant (X2(1)=6.048, p=0.014).
how do we report chi square percentages (also not sure that we need to know this)
18 of 25 (72%) of pts who received a TKA w/in the past 10 years were having contra knee pain
7 of 25 (28%) of pts who didn’t receive a TKA within the past 10 years were having contra knee pain
give % of people w/ the disease and w/ or w/o the exposure
how do we check validity of a chi square test?
check that all EXPECTED frequencies are greater than 5
if the validity check shows that some expected frequency/frequencies are less than 5, what should we do?
use the Fisher’s exact test values
are the hypotheses the same for larger contingency tables?
yes, they are still
H0: independent
Ha: not independent
what is relative risk?
how much the exposure contributes to disease
can the chi square test tell us how much the exposure contributes to disease?
nope
can relative risk be calculated with a contingency table with more than 2x2?
NO
t/f: relative risk allows us to compare probabilities in terms of their ratios
true
what does P[Y|X] mean?
the probability of Y given X
given a normal 2x2 contingency table, what is the P[disease|exposure]?
a/(a+b)
given a normal 2x2 contingency table, what is the P[disease|no exposure]?
c/(c+d)
how do you calculate the RR?
P[disease|exposure] / P[disease|no exposure]
if the RR=1, what does this mean?
the probabilities of 2 events are the same
probability of getting disease w/ or w/o exposure is the same
if RR>1, what does this mean?
there is an increased risk
if RR<1, what does this mean?
there is a decreased risk
if RR=2, what does this mean in the TKA example?
the risk for contra knee pain is 2x as likely w/TKA than w/o TKA
if RR=0.5, what does this mean in the TKA example?
the risk for contra knee pain is half as likely w/TKA than w/o TKA
how do you calculate the RR for the probability of knee pain w/TKA?
P=[KP|TKA]
how do you calculate the RR for the probability of knee pain w/o TKA?
P=[KP|no TKA]
how do you calculate the RR of having knee pain?
[KPT|TKA] / [KP|No TKA]
if the RR of the TKA example is 2.67, what is the conclusion?
the risk of having contra knee pain is 2.67 times higher for ppl who have had a TKA compared to ppl who have no had a TKA
what is the odds ratio?
another way to compare 2 probabilities
the relative likelihood of a condition (disease) given a dichotomous categorical variable
ratio of odds for 2 probabilities
t/f: odds ratio can only be used with 2x2 contingency tables
true
how are odds calculated?
disease/no disease
how is odds ratio calculated?
disease/no disease for exposed / disease/no disease for non-exposed
why do we calculate OR?
to summarize difference in likelihood of disease given risk exposure
is RR or OR the “cross ratio” of the 2x2 table?
OR
what is the equation for OR?
OR=a/b / c/d=ad/cb
if OR=1, what does this mean?
likelihood of disease is equal bw risk groups
if OR>1, what does this mean?
likelihood of disease is greater in those w/risk factor
if OR<1, what does this mean?
likelihood of disease is lower in those w/risk factor
if in the TKA example, the OR is 3.05, what is the conclusion?
the likelihood of developing knee pain is 3.05 times greater when you have had a TKA compared to when you don’t have a TKA
what is the difference bw OR and RR (don’t really have to know this)
OR is an approximation of RR
OR=RR(1-P2)/(1-P1)
t/f: the RR and OR should be similar values
true
t/f: the research question matters
true!
