Missing Data und independent variable correlated with dependent variable

[Sarah Weissmann]

Dear Statalist,


I want to find out how much the missing values of the independent variable age correlate with the dependent variable political interest.
I have made a graph that shows me that there seems to be a connection, meaning that younger people have more missings on the variable political interest. I want to know how strong the connection/correlation is between age and missings on political interest and dont know what test is approriate. I am thankful for all advice.

* Example generated by -dataex-. For more info, type help dataex
clear
input byte polinterest int(age ysurv)
.d 14 1981
.d 14 1981
.d 14 1981
.d 14 1981
.d 14 1981
.d 14 1981
.d 14 1981
.d 14 1981
.d 14 1981
.d 15 1981
 1 15 2006
.d 15 1981
 1 15 1995
 3 15 1995
.d 15 1981
end
label values polinterest polin1
label def polin1 1 "Not at all interested", modify
label def polin1 3 "Somewhat interested", modify
label values age X003
label values ysurv S020
label def S020 1981 "       1981", modify
label def S020 1995 "       1995", modify
label def S020 2006 "       2006", modify

missings.gph

[Fei Wang]

Sarah, you may simply define a missing indicator and regress the indicator on age and other covariates which you think may be associated with being missing on "polinterest". The coefficients and corresponding statistical significance may tell how relevant age and other variables are in the issue of missing.

gen missing = mi(polinterest)
reg missing age ...

[Sarah Weissmann]

Thank you!