Random Stuff that Matters — Stian Haklev

Abstraction: Practical R data wrangling tips and workflow best practices

Key points:

• knitr enables literate programming in R: mix executable code with Markdown documentation; "Knit HTML" runs from clean slate ensuring reproducibility
• RStudio server allows geographically dispersed teams to run analyses on central datasets without distributing large files to each person
• 90% of a data scientist's job is data wrangling (acquiring, cleaning, combining data sources), not running statistical functions
• ggplot2 uses Grammar of Graphics layered approach — requires learning its DSL to use effectively; data.table is faster than data.frame for large datasets with powerful split/apply/merge
• DRY principle: abstract repeated ggplot2 code blocks into functions with parameters to ease updating plot themes across many charts
• Raw data should remain unmodified; all transformations live in reproducible scripts; separate import/cleaning scripts from analysis scripts

Connections: Rstudio · Github · Data Wrangling · Literate Programming · R Programming

Source: http://reganmian.net/blog/2014/10/14/starting-data-analysiswrangling-with-r-things-i-wish-id-been-told/