You could do trial-and-error, but that would be tedious. The only other way that I can see to do this is to fit a distribution to the data, and then use the Quantile Profiler with a desirability function to find the quantile level. For the situation you proposed, I fit a smooth curve to the data and the results looked like this:
Of course different distributions will give slightly different results, but it is the only way that I can see to determine the proper quantile level.
Since you had results in a script, here is the script to match my results:
Distribution(
Continuous Distribution(
Column( :weight ),
Fit Smooth Curve(
Quantile Profiler(
1,
Confidence Intervals( 1 ),
Desirability Functions( 1 ),
Smooth Curve Quantile <<
Response Limits(
{Lower( 38.2411401370744, 0.0183 ), Middle( 115, 1 ),
Upper( 200, 0.0183 ), Goal( "Match Target" ), Importance( 1 )}
),
Term Value(
Probability(
0.685484294268159,
N Levels( 200 ),
Lock( 0 ),
Show( 1 )
)
)
)
)
),
SendToReport(
Dispatch( {"weight"}, "Quantiles", OutlineBox, {Close( 1 )} ),
Dispatch( {"weight"}, "Summary Statistics", OutlineBox, {Close( 1 )} )
)
);
Dan Obermiller