Help with Variable Name Consistency Across Multiple Files

[Jenna Kerry]

Hello,

I’m working on appending and merging several data files across different years. I’ve noticed that for the same variable, the name changes slightly depending on the year. For example, in 2020, it’s named work_hour_2020, and in 2021, it’s work_hour_2021. I’ve already managed to rename them to be consistent, but before proceeding with merging and appending to create a panel dataset, I wanted to ensure that all variables have the same names across the files.

I typically use the

 describe

command to list the variables, and I expect that if there’s still an incorrectly named variable (e.g., work_hour_2006), running:

 
des work_hour

would return an error like

  "variable work_hour not found"

alerting me that I need to rename it. However, I discovered that Stata doesn’t return an error—it instead lists the details of the original variable (e.g., work_hour_2006). This made me think that I didn’t need to rename the variables, which caused some confusion.

Is there a way to efficiently check that variable names are consistent across all the files before merging?

Also, when appending multiple files, should I rename the variable labels to ensure they correspond to the same variable across different cross-sections?

Thank you so much for any help!

[Clyde Schechter]

You can get Mark Chatfield's -precombine- command from SSC. It will enable you to identify variables that are named the same, and variables that are not, as well as identifying situations where variables named the same are nevertheless incompatible (e.g. in one data set it's a string,and in another it's numeric, or numeric in both but with different value labeling). It's a flexible tool with a fair number of options--so read the help file carefully before you use it so you can get the most out of it.

Note: -precombine- will identify file compatibility problems and alert you to them. It does not fix them--that part is up to you.

When you are appending a bunch of files, variables that represent the same thing (and are coded in the same way) across different data sets should be -rename-d to a common name so they will all end up a single variable in the final result file.

[Jenna Kerry]

Thanks for letting me know about the package. I use Stata 18.0, and it said " "precombine" not found at SSC". I guess I don't have it.

[Erik Ruzek]

Jenna - try search precombine.

[Clyde Schechter]

Sorry about that. You're right, it's not at SSC. You can get it at -net sj 15-3 dm0081-.

[daniel klein]

I typically use the

 describe

command to list the variables, and I expect that if there’s still an incorrectly named variable (e.g., work_hour_2006), running:

des work_hour

would return an error like

 "variable work_hour not found"

alerting me that I need to rename it. However, I discovered that Stata doesn’t return an error—it instead lists the details of the original variable (e.g., work_hour_2006).

Type

set varabbrev off

then try again. Stata allows abbreviated variable names by default.

Is there a way to efficiently check that variable names are consistent across all the files before merging?

describe using dataset_1 , varlist
local varlist_1 `r(varlist)'
describe using dataset_2 , varlist
local varlist_2 `r(varlist)
local only_in_1 : list varlist_1 - varlist_2
local only_in_2 : list varlist_2 - varlist_1
local in_both   : list varlist_1 & varlist_2
display "Variables only in dataset 1: `only_in_1'"
display "Variables only in dataset 2: `only_in_2'"
display "Variables in both datasets : `in_both'"