[Clyde Schechter]

You might save yourself considerable effort if you download the -zanthro- package from the Stata Journal software. (-search zanthro- and then follow the links.)

I should also add that if you have reduced your misclassification rate to 5 out of 2000 you are doing extremely well. While it feels odd to discourage the pursuit of even greater accuracy, the accuracy of weight data itself is, in most settings, not very good. So you may be trying to calculate z-score and percentiles with an apparent precision that is far greater than the underlying data itself can support. Just sayin'.

[Patrick Abi Nader]

Hi Clyde,

I already use zanthro to compute bmi z-scores and I compute percentiles out of that :-). But because I was using age in different units my percentiles were not matching the CDC ones.

Thank you for your observation on the accuracy, I am a growing statistician/ epidemiologist/ researcher and it is hard at this point in my career to know where to stop trying to improve/ ensure the accuracy of the data that I analyze. My strong ethics do not allow me to present results that were not computed with the most accurate method :-). It definitely helps to receive the perspective of a tenured researcher.

Thank you :-)

Nick I also like the way you proposed to complete the computation. Very logic!

[Patrick Abi Nader]

Quick update, Nick although your suggestion is logical it doesn't come out with what I was looking for. After I do bullet 2 in your suggestions I have odd and even numbers. Even numbers divided by 2 won't give me one month intervals at 0.5. (this is what I wanted to produce 64.5, 65.5, 66.5, 67.5).
Clyde's code does the trick. Carole's and William's approaches produce what I was looking for but for my purposes I will use the last completed .5 months as Clyde pointed out. William when I generated tblage3 and tblage4 missing data was replaced with 24 or 24.5 (those individuals did not even have age in month calculated).

Thanks again everyone!
Patrick

[William Lisowski]

William when I generated tblage3 and tblage4 missing data was replaced with 24 or 24.5 (those individuals did not even have age in month calculated)

Didn't know that your input included missing values. It's trivial to fix, so I'll leave it as an exercise for the avid Stata students among us.