As Ista and Rich have implied, graphics are better than tables for AEs. Also, statistical tests are not always of value here. For more, see http://www.ctspedia.org/do/view/CTSpedia/StatGraphHome and p. 47-53 of http://biostat.mc.vanderbilt.edu/wiki/pub/Main/StatGraphCourse/graphscourse.pdf. p. 51 contains a graphic produced by Svetlana Eden at Vanderbilt University that is particularly useful for AEs by body system and preferred term.
Frank Ista Zahn wrote > > Hi Robert, > > I think you might find it helpful to start with > http://biostat.mc.vanderbilt.edu/wiki/pub/Main/StatReport/summary.pdf > > Best, > Ista > > On Friday, February 24, 2012 07:13:10 PM Robert Wilkins wrote: >> A graph != A table. >> I'm talking about a page full of summary statistics and advanced >> statistics, with lots of cross categories on the top and left margin >> of the table, as opposed to a visual display with x-axis and y-axis, >> which is totally different. >> >> (An example of how this is done in another language is available at >> http://fivetimesfaster.blogspot.com ) >> >> For an AE table, you have an N and % column for every treatment group, >> and for all patients combined. On the right side, a categorical >> p-value (chi-sq or Fisher's) for every preferred term (every row! >> forget multiple testing issues, this is what the boss is asking >> for(it's ad-hoc safety analysis)) >> There's a row for grand total N for each group. >> A row for N and % of patients with any event (regardless of body >> system and preferred term) >> For each body system, there's a section of rows that include: >> A row for N and % of patients with any event (this body system) >> A row for N and % of patients who do NOT have an event( this body >> system) >> And , of course, within body system, a row for each preferred term >> (again N and % for each group , and also the p-value) >> >> Body system and preferred term are, of course broad medical category >> and specific medical category. >> >> >> In the Pharma industry, they use the SAS programming language. Each >> table often needs several hundred lines of code. Essentially it's a >> combination of analysis and (visual)-reporting mixed together, with >> some prerequisite data transformation. (And yes, with this new >> language, it can be done in under 20 lines of code). >> >> I have not seen people discuss attempts to do such things with the R >> programming language, and how successful such attempts have been. How >> hard is it, how much code is it? >> >> In general, we are talking about a variety of complex, >> somewhat-nonhomogeneous statistical tables with a variety of different >> row sections and row categories, and different column sections and >> column categories, and a mixture of summary statistics and advanced >> statistics (p-value , least square mean, etc), and sometimes >> statistics from different statistical procedures on the same page. >> >> Robert Wilkins >> >> ______________________________________________ >> R-help@ mailing list >> https://stat.ethz.ch/mailman/listinfo/r-help >> PLEASE do read the posting guide >> http://www.R-project.org/posting-guide.html >> and provide commented, minimal, self-contained, reproducible code. > > ______________________________________________ > R-help@ mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide > http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code. > ----- Frank Harrell Department of Biostatistics, Vanderbilt University -- View this message in context: http://r.789695.n4.nabble.com/How-would-you-program-an-Adverse-Events-statistical-table-using-R-code-tp4419007p4421608.html Sent from the R help mailing list archive at Nabble.com. ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.