Merge three datasets at the same time?

[Jaeyong Yoo]

I have three datasets (A,B, and C), which all datasets have the Year, State, Index variables. The Year and State are the same in three datasets, but the Index1 is in A, the Index2 in B, and the Index3 in C. I would like to merge three datasets.

I was able to merge three datasets with this code:

use A, clear

merge 1:1 Year State using B
drop _merge
merge 1:1 Year State using C
drop _merge

save data_merged

Is there any way I can merge multiple datasets together?

Many thanks!

[William Lisowski]

What you did is the correct approach to merging multiple datasets together. There is no single command in base Stata that will merge more than two datasets simultaneously. It is possible to imagine writing a command that would hide the sequential merges and just present the final result. I cannot imagine the circumstances under which I would find this helpful, since I generally want to check after each merge that my assumptions about matching are correct, and perhaps discard nonmatches from one or the other of the datasets.

[Leonardo Guizzetti]

If you are familiar with merging in another statistical package, I'm quite sure that the underlying process is hiding the sequential merges before filtering out the requested combination of rows. Is it more typing? Certainly. But it is more explicit for an operation that can have quite serious implications to your dataset structure and integrity.

[shem shen]

https://www.stata.com/stb/stb29/dm37/xmerge.hlp

[shem shen]

I used xmerge before. It works.

https://www.stata.com/stb/stb29/dm37/xmerge.hlp

[Leonardo Guizzetti]

I used xmerge before. It works.

https://www.stata.com/stb/stb29/dm37/xmerge.hlp

Having taken a quick read of the source code (from version 4, even) this is precisely what's happening. A better and potentially faster way would be to manage the merges with frames or using joinby or merge.

[Jaeyong Yoo]

I used xmerge before. It works.

https://www.stata.com/stb/stb29/dm37/xmerge.hlp

Thank you, Shem.

[Jaeyong Yoo]

What you did is the correct approach to merging multiple datasets together. There is no single command in base Stata that will merge more than two datasets simultaneously. It is possible to imagine writing a command that would hide the sequential merges and just present the final result. I cannot imagine the circumstances under which I would find this helpful, since I generally want to check after each merge that my assumptions about matching are correct, and perhaps discard nonmatches from one or the other of the datasets.

You're right. I might need to check each time if the merge has been correct. Thanks, William and Leonardo.