#+TITLE: Diving in: a replicable analysis
#+AUTHOR: @@latex:{\large Konrad Hinsen} \\ \vspace{0.2cm}CBM, CNRS Orléans and Synchrotron SOLEIL\\ \vspace{0.2cm} \texttt{konrad.hinsen@cnrs.fr}@@
#+OPTIONS: H:2 tags:nil
#+EXCLUDE_TAGS: noexport
#+LANGUAGE: fr
#+SELECT_TAGS: export
#+LATEX_CLASS: beamer
#+LATEX_CLASS_OPTIONS: [presentation,bigger]
#+LATEX_HEADER: \usepackage[normalem]{ulem}
#+BEAMER_HEADER: \setbeamercovered{invisible}
#+BEAMER_HEADER: \beamertemplatenavigationsymbolsempty
#+STARTUP: beamer
#+COLUMNS: %45ITEM %10BEAMER_ENV(Env) %10BEAMER_ACT(Act) %4BEAMER_COL(Col) %8BEAMER_OPT(Opt)
#+STARTUP: indent
#+PROPERTY: header-args :eval no-export


* M3-S1: What is a replicable analysis?

** 3. Diving in: a replicable analysis
 1. *What is a replicable analysis?*
 2. Case study: incidence of influenza-like illness
 3. Data import
     - Jupyter
     - RStudio
     - OrgMode
 4. Verification and inspection
     - Jupyter
     - RStudio
     - OrgMode
 5. Obtaining answers to a few questions
     - Jupyter
     - RStudio
     - OrgMode
** Traditional data analysis
[[file:img/traditional-analysis-3-svg.pdf]]
** Replicable data analysis
[[file:img/replicable-analysis-3-svg.pdf]]
** Why do it replicably?
 - Easy to re-do if the data change
 - Easy to modify
 - Easy to inspect and verify

* M3-S2: Case study: incidence of influenza-like illness

** 3. Diving in: a replicable analysis
 1. What is a replicable analysis?
 2. *Case study: incidence of influenza-like illness*
 3. Data import
     - Jupyter
     - RStudio
     - OrgMode
 4. Verification and inspection
     - Jupyter
     - RStudio
     - OrgMode
 5. Obtaining answers to a few questions
     - Jupyter
     - RStudio
     - OrgMode

** Take-home message
- No manual editing of data
- Everything is done in code!

* M3-S3A: Data import (Jupyter)
  
** 3. Diving in: a replicable analysis
 1. What is a replicable analysis?
 2. Case study: incidence of influenza-like illness
 3. *Data import*
     - *Jupyter*
     - RStudio
     - OrgMode
 4. Verification and inspection
     - Jupyter
     - RStudio
     - OrgMode
 5. Obtaining answers to a few questions
     - Jupyter
     - RStudio
     - OrgMode

** Technical choices
 - Jupyter notebook
 - Python 3 language
 - Libraries:
    - pandas
    - matplotlib
    - isoweek

** Take-home message
 - The data are read directly from the source
 - Missing data must be handled
* M3-S3B: Data import (RStudio)

** 3. Diving in: a replicable analysis
 1. What is a replicable analysis?
 2. Case study: incidence of influenza-like illness
 3. *Data import*
     - Jupyter
     - *RStudio*
     - OrgMode
 4. Verification and inspection
     - Jupyter
     - RStudio
     - OrgMode
 5. Obtaining answers to a few questions
     - Jupyter
     - RStudio
     - OrgMode

** Technical choices
 - RStudio development environment
 - R language
 - Library: parsedate

** Take-home message
 - The data are read directly from the source
 - Missing data must be handled

* M3-S3C: Data import (OrgMode)
  
** 3. Diving in: a replicable analysis
 1. What is a replicable analysis?
 2. Case study: incidence of influenza-like illness
 3. *Data import*
     - Jupyter
     - RStudio
     - *OrgMode*
 4. Verification and inspection
     - Jupyter
     - RStudio
     - OrgMode
 5. Obtaining answers to a few questions
     - Jupyter
     - RStudio
     - OrgMode

** Technical choices
 - Emacs editor + Org mode
 - Languages:
    - Python 3 for pre-processing
    - R for analysis
** Take-home message
 - The data are read directly from the source
 - Missing data must be handled

* M3-S4A/B/C: Verification and inspection

** 3. Diving in: a replicable analysis
 1. What is a replicable analysis?
 2. Case study: incidence of influenza-like illness
 3. Data import
     - Jupyter
     - RStudio
     - OrgMode
 4. *Verification and inspection*
     - *Jupyter*
     - *RStudio*
     - *OrgMode*
 5. Obtaining answers to a few questions
     - Jupyter
     - RStudio
     - OrgMode

** Take-home message
- Preprocessing
  - Adapt the data to the software's conventions
  - Facilitates the analysis
- Verify as much as possible
  - Visual inspection
  - Validation code
* M3-S5A/B/C: Obtaining answers to a few questions

** 3. Diving in: a replicable analysis
 1. What is a replicable analysis?
 2. Case study: incidence of influenza-like illness
 3. Data import
     - Jupyter
     - RStudio
     - OrgMode
 4. Verification and inspection
     - Jupyter
     - RStudio
     - OrgMode
 5. *Obtaining answers to a few questions*
     - *Jupyter*
     - *RStudio*
     - *OrgMode*

** Questions
 1. Which years have seen the strongest epidemics?
 2. What is the frequency of weak, average, and strong epidemics?
** Take-home message
 - A replicable analysis must contain *all* data processing steps in an *executable* form.
 - It is important to *explain* all choices that have an impact on the results.
 - This requires making many *technical details* explicit, because that is where most mistakes happen!