pwcorr
The same correlation grid, but each pair uses every row it can.
What we have
- the correlation matrix, each pair on the rows that have both values
Coming later
Nothing known to be missing.
How you write it
pwcorr varlist
Square brackets mean optional. A vertical bar means choose one.
What it does
Where correlate drops a row if anything is missing, pwcorr computes each pair on whatever rows have both values (pairwise deletion). You keep more data, at the cost of different cells resting on different people.
Which to prefer depends on why the values are missing. When in doubt, run both: if they disagree noticeably, the missing data is telling you something.
Try it
Every example below was run after sysuse vira.sample02, clear.
. pwcorr mpce age education
| mpce age education
-------------+---------------------------------
mpce | 1.0000
age | 0.0683 1.0000
education | 0.2803 -0.0329 1.0000
The same thing in pandas and in R
pandas
df[['mpce', 'age', 'education']].corr() # pairwise
R
cor(df[, c("mpce", "age", "education")], use = "pairwise.complete.obs")
# This R was run in a real R (4.6.1) and gave the same numbers.