Exact p-value reported as missing of Kolmogorov–Smirnov test

[Nate Liu]

Hello,

When I was conducting the Kolmogorov–Smirnov (KS) test, the exact p-value is reported as missing with the command below (where the exact option is specified).

ksmirnov start_year, by(original) exact

where start_year is the start year of a given event and original is the classification variable (1,090 observations have a value of 1 and 2,126 observations have a value of 0).



The result below shows that the exact p-value (in the last column) is reported as missing.





I was thinking it might be because the year variable is not continuous enough so I also tried a different variable which is more continuous but the exact p-value is still reported as missing.



Could you please let me know what might be the possible reasons for that?



In addition, when I used the following the command to draw the histogram (with the start year variable), I got the histogram but not quite sure what the density means on the y-axis. I was thinking it might be p.d.f but the sum clearly exceeds 1 in either situation (i.e., classification variable = 1 or 0).

histogram start_year if original == 1, ylabel(, angle(0)) xtitle("original1") name(fig1, replace)
histogram start_year if original == 0, ylabel(, angle(0)) xtitle("original0") name(fig2, replace)
graph combine fig1 fig2

Thank you very much for your help!

Best Wishes,
Nate