Common pitfalls

• build

  • Not mapping the netCDF before extracting data from it
  • Common Stata mistakes
  • Top ten Stata gotchas
  • using encode to define unique IDs in different datasets before merging them
  • not using assert to verify conditions as you prepare the data (especially useful right after merge)

• analysis

  • watch out for the N: missing values in 'control' variables i.e. other than the outcome or primary predictor could cause the estimating sample to change

• other tips (these aren't mistakes as such)

  • not ensuring that unique IDs have a fixed length (pro tip: you can define new string variables which embed numeric codes with leading zeros gen newvar = string(var1,"%02.0f") + string(var2, "%03.0f"). here's an example of how you can use assert to verify that coded the new variable correctly:

gen newvar = string(var1,"%02.0f") + string(var2, "%03.0f")
assert length(newvar ) == 5