rreport To Do List

1. TERRI accrualReport:
   • Have made several improvements:
     1. Modified order of generated plots and tables.
     2. 'Subjects randomized over time' plot:
        1. Can now specify (up to) two EntryDate variables.
        2. xlim has been modified to the 'floor' of the minimum date to the 'ceiling' of the maximum date are used - i.e., if 4/16/06-6/20/06 specified as dateRange, then xlim=c(4/1/06, 6/30/06).
        3. On the date (x-) axis, 'minor' (i.e., shorter) tick marks are used to denote each month, and 'major' (i.e., longer) tick marks are used to denoted every six months, which are labeled.
     3. Output of other tables and plots depends on whether just Minor or Minor and Major are specified.
        • If just Minor specified, then 'Number of subjects randomized by Minor' plot and 'Number of Minors having a given number of subjects randomized' plot are generated.
        • If Minor and Major specified, then 'Number of subjects randomized by Major' plot (with number of Minors within each Major given), 'Number of subjects randomized by site' plot (automatically implements combineEqual if greater than 40 levels; and adds the number of subjects per site if combineEqual), and 'Number of sites having a given number of subjects randomized' plot are generated. Also generates table of combineEqual abbreviations. The definition of 'site' is given in each appropriate table caption.
2. TERRI rangeCheck:
   • Have split into two separate functions: rangeCheck and makeNA.
   • rangeCheck creates a LaTeX table that displays the frequency of various lab parameter values outside a defined range. The table has three columns: (1) Lab parameter; (2) Defined range; and (3) 'Out of range' (the percent and frequency of values outside the defined range).
     • Improvements to make: Have rangeCheck create a second table (and/or dot plot) that summarized the sites (country and center) (and IDs) that have records outside the defined range of each lab parameter.
   • makeNA replaces the values of a data frame that are outside the defined range with NA.
3. aeReport2:
   1. THOMAS Allow to have short captions for table of contents.
   2. Develop a plot method.
   3. THOMAS Allow to generate a title and table headings even if the table will be blank (i.e., nrow(data) = 0).
   4. THOMAS Make it easy to direct the output to a cumulative appendix that has multiple sections. Modify to have an option to put in the body of the report only those events having a combined frequency of at least minfreq. When minfreq is used, the existing long table should go to an appendix. Create a sentence that links to the appendix (both in text and as a hyperlink).
      • See listTable item.
4. listTable:
   1. SVETLANA For functions that create listings of individual patient records, make it easy to direct the output to a cumulative appendix that has multiple sections. Default for listTable will go appendix but user can override this for special listings such as severe events of a certain type. Text written to the appendix can keep append ing to gentex/app.tex. Will need to generate \section commands automatically.
   2. THOMAS Move functionality into latex.data.frame.
5. Dot charts:
   • THOMAS Possible solution: include dot charts in table output via a dotchart=TRUE argument.
   1. Add a new mincat option to appropriate functions. Dot charts having fewer than mincat categories depicted in the entire chart should be suppressed. A good value of mincat might be 7.
   2. See if producing either a dot chart for categorical variables or a table but not both will work.
   3. Dot charts should have category frequencies added in the right margin and P-values in the (probably) left margin; idea is to have as much information as possible in graphs, then to see if committee members can dispense with tables; will have to see which details are OK to omit from a graph (e.g. treatment sample sizes as these are usually almost equal).
6. Hmisc:
   1. Submit floor.chron and ceiling.chron to the chron maintainer for consideration of inclusion in chron.
   2. THOMAS Create a new function scaleReport or adapt mixedvarReport and other functions to report discrete-value scale variables (such as those on an integer 1-5 scale) as mean and SD and with a dot chart showing all unique values of the scale. Don't use quantiles, box plots, ECDFs, or binned histograms for scales. Better: define function discreteNum that can be used when defining a data frame or in a call to summary.formula, e.g. summary(treat ~ age + discreteNum(painSeverity)). discreteNum implies you the way to compute descriptive stats is to not use quantiles but to use mean, SD, and occasionally mode.
   3. THOMAS Improve dotchart2 to switch to smaller font for left margin labels or to word-wrap them, to avoid squeezing dot charts to be too narrow. Specify a minimum width allowed?
   4. Be able to define different number of lines for the first page and then subsequent pages in a longtable = TRUE argument call.
   5. Fix units to always be placed in the right of the column in a mixedvarReport call -- \hfill in LaTeX doesn't always work.
   6. Be able to break the table into mutliple tables based on categorical "response" variable in a method = "reverse" (i.e., the categorical variable has 6 levels, so break the table up into three tables each with 2 levels of the response).
7. Confidence intervals for differences in two proportions
   • Wherever such confidence intervals of their half-widths are calculated, including in the new stuff Thomas is implemented with micro plots inside tables, use the Newcombe method. This is the default in the ci.pd function in the Epi package, which should be a required package for rreport . Note however that the confidence intervals from ci.pd are not symmetric around the difference in proportions, so for now let's use the simple normal approximation.
8. Other function specific:
   1. SVETLANA Change mixedvarReport and labreport to use new defaults:
      • in open reports, suppress ECDFs and just show histograms
      • in closed reports, choose ECDF if two treatments, replace this with box-percentile plot if > 2 treatments.
      • add an option of plotting contin. variables (not histograms) contDataPlot =TRUE.
   2. SVETLANA design and code automatic latex markers management.
   3. SVETLANA Develop a new function for comparing a data frame with the previous version; extremely concise (more concise than describe ); output is for analysts in statistical center and sponsor; need to have an ID variable so can compute number of subjects added/removed and count number of unique subjects. Should be added to Hmisc for one pair but a superfunction added to rreport to do this automatically a series of data frames.
   4. Look into replacing global variables (such as graphics output option) with something like option( ) or .Options$element.
   5. Need visit-oriented mixedvarReport.
   6. TERRI Improve put2web to put in the body of the e-mail which pages to print. Have rreport tell LaTeX to save the page number of the start of the appendix.
      • See aeReport2 and listTable appendix related items.
   7. THOMAS TERRI publishPdf 'scp -p' problem and permission problems in general:
      • Need to be able to upload and to have real date/time stamps.
      • Using Subversion solves premission problems and date/time stamping as well as providing and archive of previous report versions.
9. rreport in general:
   1. TERRI/SVETLANA Change "enrollment" to "screened" everywhere in Oreport and report .tex files.
   2. TERRI Create Rd help files for all existing rreport functions.
   3. Develop a procedure for validating pivotal calculations against the blinded sponsor's calculations by creating dummy treatment assignments that are given to the sponsor.
   4. Make plans for an automatic validation process by having rreport generate Stata code to repeat analyses.
   5. Possible to consolidate the unique number of functions in rreport?
   6. Make directory structure more user determined -- fileNames.s.
   7. Construct Makefiles for mutliple-step procedures (suggested in Frank's email).
      • That is, split report.s into datamanip.report.s and generateoutput.report.s.
   8. Incorporate ggplot package.
   9. FIXED arg append=TRUE in mixedvarReport possibly causes outputted images to overwrite each other.
   10. Table 34, 38, 39, 43 (REDUCE, open report, section 12.1,Occurance of HGPIN for subjects diagnosed with cancer, table 34): put into scientific notation in LaTeX? (improve latex.summary.formula.reverse) - ask Frank what he meant.
   11. Make mixedvarReport() plot only categorical variables with number of levels less than three when argument 'categDataPlot' equals TRUE.(Frank - REDUCE)
   12. Make ecdf() and mixedvarReport() plot the data in a given range - given xlim. (Frank - REDUCE)