home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Shareware Overload Trio 2
/
Shareware Overload Trio Volume 2 (Chestnut CD-ROM).ISO
/
dir26
/
epi601_2.zip
/
FILES07.EXE
/
EPITUTOR.PGM
< prev
next >
Wrap
Text File
|
1994-08-22
|
5KB
|
111 lines
ECHO The program (CASECON6.PGM) that just ran does a lot of
ECHO setting up of variables so that you can process them, but
ECHO it does not actually produce output. You can now give
ECHO other commands like TABLES and MEANS from the keyboard to
ECHO explore the relationships.
ECHO
ECHO If you would like a printed copy of the results,
ECHO type ROUTE PRINTER or press <F5> now. Otherwise, press <Enter>.
ECHO ?Epi> ?
ECHO
ECHO To display or print the contents of CASECON.PGM, use
ECHO TYPE CASECON.PGM (press <Enter> to skip).
?Epi> ?
ECHO You will want to check the results of CASECON.PGM by
ECHO LISTing selected variables for cases and controls.
ECHO It is easier to see the results if you sort them first.
ECHO Type SORT GROUP IDNUMBER
?Epi> ?
ECHO We will do LIST GROUP CLAMS CLAMSOUP ANYCLAMS for you.
ECHO Press <Enter>
?Epi> ?
LIST GROUP CLAMS CLAMSOUP ANYCLAMS
ECHO How comparable are the two groups? You might compare them
ECHO with respect to AGE and WEIGHT, two continuous variables, as
ECHO follows:
ECHO
ECHO EPI>MEANS AGE GROUP
ECHO EPI>MEANS WEIGHT GROUP
ECHO Here are the results:
MEANS AGE GROUP
MEANS WEIGHT GROUP
cls
ECHO Now comes the exciting part--comparing the cases and controls
ECHO with respect to food consumption:
ECHO
ECHO Type TABLES < a food variable>) GROUP
ECHO To see the names of the food variables, press <F3>
?Epi> ?
ECHO
ECHO Now do TABLES CLAMS GROUP
?Epi> ?
ECHO How about that for an odds ratio??!!!!!
ECHO
ECHO <Enter to continue...>
?Epi> ?
cls
ECHO
ECHO So far you have obtained some very large odds ratios for
ECHO CLAMS and CLAMSOUP, but if something toxic is contained in
ECHO the clams, the appropriate tabulation would be for
ECHO consumption of any kind of clams. We have set up a variable
ECHO in CASECON.PGM called ANYCLAMS, which is "+" if either CLAMS
ECHO or CLAMSOUP is "Y". The "+" is used instead of "Y" because
ECHO "+" comes before "-" in the computer's alphabet, and "Y" does
ECHO not come before "N." This sets up the order of things in the
ECHO tables so that they follow conventional epidemiologic format
ECHO with both risk factor and disease presence in the upper left
ECHO corner and their absence in the lower right corner of the
ECHO table.
ECHO
ECHO Now let's make tables of the defined variables:
ECHO
ECHO EPI>TABLES ANYCLAMS GROUP
ECHO EPI>TABLES ANYLOBSTER GROUP
ECHO
TABLES ANYCLAMS GROUP
TABLES ANYLOBSTER GROUP
ECHO
ECHO The odds ratio of over 400 obtained for ANYCLAMS is truly
ECHO astronomical when compared with the odds ratios of 2 or 3,
ECHO that can be of significance in chronic disease studies. The
ECHO relative risks given for each table should be ignored because
ECHO this is a case-control study, and the size of the exposed
ECHO population is not known. Can you think of any reason why
ECHO the odds ratios for some foods show negative ("protective")
ECHO associations, having odds ratios less than one?
<Enter> to continue..
?<Epi> ?
cls
ECHO Despite the striking results obtained so far, we should look
ECHO for explanations other than the simple hypothesis of a clam-
ECHO associated toxin or microorganism. Perhaps water consumption
ECHO is linked either positively or negatively to clam soup
ECHO consumption, and also to disease incidence. Water
ECHO consumption would then be a confounding factor. We can test
ECHO for confounding in the analysis by stratifying on WATER, that
ECHO is, by making separate tables for water consumers and
ECHO nonconsumers and by combining the results in a special way
ECHO called the Mantel-Haenszel stratified analysis. To do this
ECHO for CLAMSOUP, you would type:
ECHO
ECHO EPI>TABLES CLAMSOUP GROUP WATER Press <Enter> to do it
?Epi> ?
CLS TABLES CLAMSOUP GROUP WATER
ECHO
ECHO
ECHO Try the same approach with other possible confounders until
ECHO you are satisfied that the analysis is sound. By stratifying
ECHO on a variable, we eliminate the effect of confounding by that
ECHO variable. The approach is only valid, however, if the odds
ECHO ratios in individual strata are similar in value. If the p-
ECHO value of the chi square test for interaction is less than 0.05,
ECHO this indicates a significant difference between Odds ratios in
ECHO different strata so that the Mantel-Haenszel summary odds ratio
ECHO should be interpreted cautiously.
ECHO
ECHO You are on your own now with this wealth of variables. When
ECHO done, press <F10> to return to the EPI6 menu.
ECHO