tihs unoseqit mkesa me anwt to eat an e iocl coekio and opeh i bdlee tou
iDd no eon etinco atth the ddsO itroa on het opt etfl si rng?ow mA I niismsg nmhti?seog If ouy aealtccul ,ti sit' 6 usjt ekli eth otp rgiht .oe.n..
OR gt;&1 tcsdeiian iaecnsedr cnrcercoeu of netev. ehT olyn OR rraeteg than 1 saw ni het etlba ttha adcinidte tath teh betscuj eat isoocek btu di'tnd nkird lkm.i su,Th tath is eth olny noe wtih a taginincsfi oerncrceuc
For a more systematic approach. First look at cookies p-val is sig when not stratified, the top table is stratified the OR > 1 => sig => cookies have association.
Then look at milk p-val is sig when not stratified, the bottom table stratified the OR = 1 => loss of significance => milk have no association.
Uworld ID 1173 has a good explanation for how to look at stratified analysis.
eTh tfac taht eth dsod otria in het otp ltfe si otrcecirn amske hsit nteoiusq yrev iictlfufd. tI kmesa it eaaprp as if eht coskoei rae stcveauai tbu eht kmil had eoms otveicterp rato.cf So .ouosxiobn
Ilatilnyi ilmk giinrdkn saw dssoatieac thwi l.icEo ekrbouta whit 3.RO=9 adn .;tl0001P& t.Sifgcnin.(i.a) rfteA itnstartfaioic iotn ate iesckoo dna did ton eta eiscoko OR ecmaeb 1 iaetsdn of .39 gneiamn teh isaiocnatso eipapdrsea.d o,ereThref igtane oocekis swa a odnnceorfu nad rheet si no alre niaoaioctss tweebne nindrgik klmi nad .o.ia.ein.tdsEc,l. ilskm' e(ht eofunnrodc) bnuitoicntor swa snpsrbeielo fro eht OR fo 9.3 ni the sfrit .lecpa hsTi wsa uherderft asdndetoemrt hwti RO of 6 in teh csokoei elaon .orpgu
For people who generally had trouble reading the two charts:
First chart: We separated the entire population into two smaller populations to test for the cookies affect. In Population A (drank milk) there was an odds ratio of 6 (typo in the actual chart). In Population B (did not drink milk) there was an odds ratio of 6. Since the odds ratios are not 1, we can conclude that the cookies have an effect regardless of the population (ie drank milk people versus didn't drink milk people).
Second chart: New set of populations to test for the effect of milk. In Population C (ate cookies) there was an odds ratio of 1. In Population D (did not eat cookies) there was also an odds ratio of 1. This means that milk did not have an effect ever and didn't contribute to the disease.
"Only cookies are independently associated with E. coli cases" means that only the cookies cause the disease without the effects of something else.
siTh eno three were rufo odds oir,tsa eno ddporevi nedru hace .eblta eTh lyno eon atht had na ddso iator argeret than 1.0 was the aeblt ni eht opt htgir (
dsdO aRtoi = 6, I blee,evi) hhiwc whne uoy koedol ta eth a,sbell led to the rgtih nrewsa.
"An dosd artoi of 1 tcdisnaei hatt eht tniodocni or teenv eunrd utsyd is ayleqlu elliky ot ouccr ni obht .osupgr An dosd oatir eargrte tnha 1 dentciisa tath eth tinincood or veetn is mroe leilyk to ccuro in eht sitrf upr.o"g (eir/wnse:.O_k/aptkiir/so/tagddphdioiiwt.)
The OR in the upper left 22 table is incorrect, which should be 6 (726/36*2 =6), not 1. This means the OR of "ate cookies" does not change after stratification by "drank milk", so "drank milk" is not a confounder, and "ate cookies" is independently asso w/ EHEC outbreak.
On the other hand, OR for "drank milk" changed a lot (from 3.9 to 1.0), which indicates "drank milk" might be a confounder and, therefore, is not independently asso w/ EHEc outbreak.
we can make things simple like this way: if we want to know whether X1,or X2 correlates Y, we just separately test X1 and Y, and X2 and Y accordingly. When test X1 with Y, we require no X2 exposure; When test X2 with Y, we require no X1 exposure;
We test cookie with diarrhea, when milk was not drunk (top right): positive We test milk with diarrhea, when no cookie was eaten (lower right): negative
conclusion: only cookie correlates to the diarrhea
A question i more generally have is...
Is it possible that when you stratify the data, (i.e. comparing the effect that eating cookies has, in people who drink milk or people who do not drink milk) that the odds ratio will show significance for one but not the other?
Said differently, in the example above, could eating cookies in people who drink milk lead to a significant increase the risk of infection, but not in people who who didn't drink milk?
I've looked around in this comment thread, and have seen people mention the term "effect modification"; is that what my example above would show?
In other words, if eating cookies + drinking milk leads to a significant risk, but eating cookies + not drinking milk has no associated risk, would that mean that the milk has an "effect modification" on the risk of getting infection in people who eat cookies?
The eokydwr si NoILiYseD"cNEsaTNEPad(D.tE)" hWich in anmuh gegaulna manes T"NO OSSAEI"AC.DT