Skip to content

Latest commit

 

History

History
44 lines (32 loc) · 2.94 KB

differences-from-Stata.md

File metadata and controls

44 lines (32 loc) · 2.94 KB
Raw

Differences between Douglass.jl and Stata

General

  • Keep in mind that Douglass.jl only takes your command, parses it, and calls the appropriate functions in DataFramesMeta.jl or DataFrames.jl. Hence, you are working with the types in Julia (Int64, Float64, etc.), and the results of your operations are the results of the operations in Julia. This applies, in particular, to the behavior of missing and to the rules on type conversion and promotion.
  • Variables in the activate DataFrame are referenced using colons (:), e.g. :myvariable. If you do not use the colon, Douglass assumes that it's a variable in the current scope, e.g.
> x = 1.0
> d"gen :x = x"

creates a new variable that is being assigned the scalar 1.0 in every row.

  • Macros: Stata's local and global macros are not supported. You can create strings in Julia and "interpolate" into the strings that you send to Douglass using a double dollar sign $$, e.g.
> name_of_new_var = "myname"
> d"gen :$$(name_of_new_var) = x"

The current implementation of this rests on Julia's eval(), however, and is therefore slow (so it's not real string interpolation!). If you know of a better solution, please let me know.

  • You cannot use abbreviations for variables like in Stata (this is bad practice anyway), and in general you cannot use abbreviations for commands (exceptions to this are gen for generate, and rep for replace).

Specific to certain commands

generate and replace

  • generate and replace operate on scalars of the current row, i.e. :x refers to :x[_n] where _n is the current row index (in fact, internally Douglass.jl is filling in those indices if you do not specify them explicitly).
  • generate currently does not support Stata's explicit type declarations of the form generate long :x = 1. You can, however, use Julia functions to make sure that your assigned value are of a particular type, e.g. generate :x = Float64(1).

egen and ereplace

  • Because it's a vector-valued operation, _n cannot be used.

reshape

  • Note that the command is called reshape_wide and reshape_long here, with a _ instead of a whitespace.
  • Unlike in Stata, @ cannot be used as a whitecard.
  • In reshape_long:
    • Unlike in Stata, the variable in j() is made to have type Symbol and is not automatically converted to String or numerical types.
    • Unlike in Stata, variables that are not listed in the arguments, i, or j are not part out the returned DataFrame. If you want to include them, list them as part of i.

merge (merge_11, merge_1m,, merge_m1)

  • Note that the commands are e.g. merge_m1 not merge m:1. There is an underscore and the colon is missing.
  • The commands currently do not produce any variable _merge and no summary statistics.
  • m:m merges are currently not allowed.