STM Package: Including Year from direct Quanteda DFM intake

Question

I am using the stm & quanteda packages. My dfm was created with Quanteda and contains the Variable "Year" as document variable too. Using the stm() function, I'm able to directly input the dfm to the function, as described in the vignette. I do not understand though, how I could include the year in the stm() function. Put shortly, I'd like to include "Year" as seamlessly as the DFM in the stm() function. Please suggest a solution. My code is not available. Thank you!

Answer 1

If you are using stm, the variable year must be incorporated using the s() function, since it is a temporary variable that the model needs to calculate the coefficients. Here is an example of how to incorporate "Year" into an stm model, where the number of k = 30. In this case, the variable year can be combined with other variables in your data, such as in this case "Country", but it's not mandatory.

model_stm <- stm(documents = docs_model,
                 vocab = vocab_model, 
                 prevalence =~ country + s(year), 
                 K=30, 
                 data=meta_model, 
                 init.type = "Spectral", 
                 seed = 123, 
                 max.em.its = 500)