# R Markdown workflow

You have seen three basic workflows:

1.  Work in console.
1.  Work with a single script.
1.  Work with multiple scripts in a project.

R Markdown introduces a four workflow: instead of writing R scripts in the script editor, you write R Markdown documents. The chief advantage of this workflow is it put prose and code on an equal footing. That is obviously important if you're writing up a report to share with someone else. But it's equally important when you're working on a data analysis that's just for you because it provides a great environment to develop you code, and record your thoughts as they occur. It allows you to a modern day lab notebook --- you record everything you tried as you try it. You notebooks only need to be understandable by you. 

Knitting an `.Rmd` is also great reproduciblility practice because every time you knit it, it's done in a completely fresh session. If you always finish the day by knitting your `.Rmd` and making sure it works, then you're doing reproducible research!

R Markdown provides a useful way to organize your data science projects. You can use an R Markdown file to create a reproducible record of how you import, tidy, transform, visualise, and model data.

and then generate a report from your .Rmd file to communicate your results. This process is very efficient: you can write your report once and then deploy it many times in different formats or with different parameters. Moreover, by creating an R Markdown file, you participate in two movements that are leading to better scientific practices: 

1.   **Reproducible Research** - The familiar scientific report format 
     (*Introduction*, *Methods and Materials*, *Results*, *Discussion* and 
     *Conclusion*) helps experimental scientists report their results in a 
     reproducible way. Embedded in the format are the details that a scientist 
     would need to reproduce the experiment. 

    R Markdown files provide the same service for data science. Embedded in 
    the file is the code that a data scientist would need to repeat your analysis. 

2.   **Literate Data Analysis** - like 
    [literate programming](https://en.wikipedia.org/wiki/Literate_programming),
    literate data analysis intermingles code with human-readable text to build 
    a program that is easy to understand and easy to debug (and often more 
    streamlined than a program written in a non-literate way).