Stata in Quarto

Here, I am following the guidance of Fernando Rios-Avila blog post Using Quarto for Stata dynamic documents. It’s a simple solution that requires a local installation of Stata. Crucially, it must be Stata 17 or higher.

The approach uses a Python package called nbstata. So, if you are compiling your Quarto document (website or e-book) within RStudio, like myself, it feels a bit like running Stata via Python via R.

For those with an older version of Stata, take a look at this alternative approach which uses the stata_kernel Python package.

Installation

You need to ensure that you have a local installation of Python. I am using Anaconda. Next, you need to install nbstata. You can do this using the Anaconda Prompt

pip install nbstata
python -m nbstata.install 

You will also need to check that R is aware of the the Python environment you are using. I did this using the reticulate package. In your RStudio console, check

library(reticulate)
py_config()

This should verify the location of python.exe on your system. If not, RStudio is not aware of your Python environment. Go into Tools > Global Options… > Python. Then Select > Conda Environments and look for your python.exe. Select and then Apply changes.

Next, check that Quarto has has python3 and nbstata listed as kernels. In the RStudio console you can type:

system("quarto check jupyter")

Alternatively, in the RStudio Terminal type:

quarto check jupyter

You should see listed: “Kernels: nbstata, python3”. If not, you need to troubleshoot a while longer. AI can be very helpful for this.

Document Setup

With everything installed, you need to include the following in the YAML of your .qmd file. You can also include in the _quarto.yml file, in which case all documents will take on this setting.

---
jupyter: nbstata 
---

Codeblocks

Now you are ready to add Stata codeblocks to your Quarto markdown file. You do this as follows:

```{stata}
describe
```

The standard execution options for Quarto codeblocks work just as they do for R. The only difference is that you use the indicator *| in place of #|, because in Stata * is used to denote a comment.

A very nice feature of this package is that the Stata code is compiled together, so objects defined in one codeblock are visible to another. For example,

local test "Am I still defined in Stata?"

Running C:\Users\ndl1\OneDrive - University of St Andrews\Documents\Resources\E
> xperiments\experiment-1\profile.do ...

Now in a new codeblock:

display "`test' Yes"

Am I still defined in Stata? Yes

Note, this works for Stata locals which are not defined across do-files in Stata. This is a nice feature of the nbstata package.

Limitations

This approach to incorporating Stata in Quarto compiles the document using the nbstata kernel. Unfortunately, this does not allow you to then have codeblocks in other languages. I will demonstrate an alternative, but slightly more complicated, solution to this later on. If you only intend to use Stata, this is the neatest implementation of Stata.

If you are compiling a website or e-book with multiple .qmd files you can always execute Stata using nbstata in one file, while running R and/or Python in others. The issue is with multilingual files.