class: center, middle, inverse, title-slide # R or Python: ## A Programer’s Response ### Jiangtang(‘JT’) Hu <br/> d-Wise ### 2018-06-05 --- class: inverse, center, middle # Disclaimer # 前方高能 --- # A Personal Journey to SAS, R and Python * Started to learn SAS at graduate school in 2006 * I then worked for SAS, Sanofi and d-Wise * Blogging at http://jiangtanghu.com/blog/ --- # A Personal Journey to SAS, R and Python * Start to learn R in 2009 * Still a frequent R user + blogging + slides + course works + sometimes in professional work --- # A Personal Journey to SAS, R and Python * Start to learn Python in 2014 * Still a learner! + MOOCs (machine learning/deep learning) + pioneering work (deep learning in Clinical world?) + occasionally in professional work --- # Why Tool Matters * You are what you touch + Tool is a part of your body -- * Path-dependence + Think about SAS Version 5 Transport File!  --- # R is comming to clinical - R is well known in statistical departments - Long time language battle between SAS and R - Some biostatisticians in pharma use R - Some FDA statistical reviewers use R --- # Python > Python is not a statistic computing language! > and this may be its greatest strength as a language for statistical computing. ^* .footnote[[*] Python's Data Science Stack by Jake VanderPlas, JSM 2016 ] --- # Language Design Python is simply a better language. --- # Language Design: R Ross Ihaka: problems with current R systems ^* * Tree-walking interpreters: no optimisation, inherently slow * Call-by-value semantic: Produces vast amounts of data copying + the design matrix is copied 6 times during the linear model fitting process .footnote[[*] Ross Ihaka: Towards a New Statistical Computing System, JSM 2014 ] --- # Language Design: R Solution #1: Radical * Ross Ihaka: new software! ^* .footnote[[*] Ross Ihaka: Towards a New Statistical Computing System, JSM 2014 ] --- # Language Design: R Solution #2: system-wise modification * renjin, is a JVM based R interpreter (developed by Alexander Bertram) * Microsoft R Open * TIBCO Enterprise Runtime for R * ... --- # Language Design: R Solution #3: package-wise optimization * Hadley Wickham: tidyverse --- # Language Design: Python Python is getting better! * Text and binary data in Python 2 are a mess * Python 3 started at 2008 + backwards-incompatible change, dangerous move! * Python 3 is widely accepted * Python 2 will end at Jan 1, 2020 --- # Ecosystem: R Although facing fundamental drawbacks in core design, R enjoys huge success due to its user contributed packages. --- # Ecosystem: R One of the problems of R ecosystem is most packages are one-person-project * Hadley Wickham: tidyverse -- + https://github.com/tidyverse/ggplot2 + https://github.com/matplotlib/matplotlib --- # Ecosystem: R One of the problems of R ecosystem is most packages are (pure, isolated) collection of functions * E1071 --- # Ecosystem: Python  --- # System Integration -- R is a free software environment for statistical computing and graphics. -- Python is a general-purpose programming language: * data sciences * web frameworks * system administration * networking, automation, databases, ... -- --- # Back to the Future Python is the first-class citizen in the world of * big data * artificial intelligence * machine learning * deep learning * ... --- # Kaggle: The State of ML and Data Science 2017 ^* Python was the most commonly used data analysis tool across employed data scientists overall, but more Statisticians are still loyal to R.  .footnote[[*] https://www.kaggle.com/surveys/2017 ] --- class: inverse, center, middle # Thanks! ^* Jiangtang('JT') Hu d-Wise jiangtanghu.com/blog