Why do I use wiki for research?

• Multiple projects
  • Involve with many research projects concurrently.
  • It's hard to manage research activities across different projects.

• Time issue
  • Research takes time.
  • Clients often visit my office after many months of inactivity. This means:
    • I may not remember what I did 3 months ago on a particular project.
    • "What did I do before?" "How was this variable defined?"

• Create an analytic summary before presenting to clients in a meeting.
• Tables prepared in Excel. Plots in a graphical format. Combine them & write analysis descriptions in Word.
  • When on a tight deadline, even used raw computer output.

• You end up with too many summary files:
  • e.g., Analysis 03122019 v3.docx; Logistic model adj Kcal sleep 12022018.xlsx
  • Can you tell what analysis was done in which file?
• Keep sending files to others via email – Recipe for disaster!

Did you receive this?

No I didn't. Send it again

Sorry I forgot to attach this file!

See this file for this analysis & see that file for that analysis.

I'm on zoom. What table are you looking at now?

Can you include this variable to this table & send me again today?

See attached. This is version 3 of today's summary!

• Imagine you have to dig through these emails, looking for results that were revised multiple times after many months.
• Or maybe you end up with a pile of printouts, hopelessly looking for the "right" table.

• Need a better way to present results to others!
• Need a better way to organize research activities!!

• No more searching for files and output. No more sending them to others.
  • Everything is in one place
    • "See my wiki page online. I just added more results a few minutes ago."

• Easy to create a page and post results there.
  • Include tables & figures. Can post as files or graphics.
    • "The table is posted in my wiki. You can download it in an Excel file if you want."
    • "Journal is asking graphs in a vector file? Just get the PDF file from my wiki."

• Hyperlinks
  • Found a relevant paper that you want to share with others? Need to include it in references later? Just insert a link.
  • Cross-reference between wiki pages.

• Collaborative editing
  • Most of research require collaborative efforts. You don't know the whole aspect of a research project.
    • You need someone else to provide some specific information on a research.
    • "Can you fill in this information in my wiki, please?"
  • Get feedback.
    • Your collaborators can insert comments and add discussions.

Must assume that the future you won't remember anything about the project!

• Study design
• Subjects
  • Recruitment, inclusion/exclusion criteria
• Datasets:
  • How many data files are involved?
  • When data files were received? When data files were updated?
  • Outcomes, variables, the number of obs

• Data integrity
  • Error checking and validation
• Data management
  • How multiple datasets were merged?
  • How many subjects were eventually used?
  • Derived variables and how they were defined
    • If necessary, include your code/syntax in wiki
    • Can include equations using LaTeX notations, like $L(\boldsymbol{\theta})= \prod_{i=1}^n f(y_i|\boldsymbol{\theta})$
• Demographic tables, descriptive statistics
• Descriptions of statistical models
• Model building/specification, model diagnostics
• Supplementary analyses

• Include lots of figures
  • Figures are extremely useful especially during data checking and exploratory phases.
  • Most of these figures won't be used in a paper, but that's okay.
  • Embed figures in wiki pages. Need to post graphic files first.
• Tables
  • Most of the time I simply copy and paste computer output, with minimal editing
    • This makes it easy to revise tables upon request (which happens always)
    • But do add titles, superscripts and footnotes to make them self-explanatory
  • For large tables, create in Excel and post the file
• Results
  • Write brief interpretations of results
  • Comment on results that you find interesting or unexpected
  • Detailed output can be included, but hidden in a page (folding/unfolding texts)
• Update results as needed
  • Can always revert back to previous edits
  • If necessary, keep old results in a separate wiki page and create a link to it
• Present results in a logical order
  • Do NOT present results by date

• There will be many meetings during a project. Keep meeting logs.
  • Do this by date, the latest date on top of the list
• Meeting logs may include:
  • Attendants
  • To do list and reminders
  • Change of plan
  • Next meeting date
• Keep the log to CYA in case something goes wrong

• Can check/edit wiki pages anytime, anywhere
• Analysis on the go using R and RStudio in a USB drive

• Nested hierarchy of contents
  • One front page for each research project
  • Can create multiple wiki pages branching from there
  • Not so easy in Google Docs
• A wiki page has different levels of headlines
  • TOC automatically generated on a page based on headlines
• Forces to think a logical way to present results
• Preserve page history
  • Can revert back to old revisions if necessary

• Reproduciable research: Make your analysis more transparent
  • If necessary, copy & paste code/syntax used for analysis

• For security purposes, may need to prepare a user account for every collaborator.
  • Requires a password
  • Only authorized users can view/edit certain wiki pages.
• DokuWiki has an easy way to manage user groups and access control.
  • I usually create a user group when a new project starts.

• Can post results, but do NOT post data files.
  • Electric data should be securely stored in relational databases, such as a SQL server.
  • Users should access data directly from a SQL server or other secure data storages with password protection.
  • Avoid file transfer between users as much as possible. It is the weakest link in data security.