Home Examples Learning More FAQ Bibliography Glossary Conferences Software Resources

Basis
Functions

Goods

Growth

-What's wrong with fitting the data exactly?
-How do I actually do this?
-How do I actually do this? (cont.)

Weather

Valid XHTML 1.0!

Smoothing the Growth Data
Previous - 1 - 2 - 3 - Next

Expertise: Beginner

How do I actually do this? - Continued
The results are much more interesting when we look at the velocity curves after smoothing in Figure 3. The Matlab command to do this is

plot(hgtffd, int2Lfd(1))


Figure 3: The height velocities of the first 10 girls in the Berkeley Growth Study
after penalizing roughness at the level of the fourth derivative.

Now the velocities seem acceptably smooth, and we can easily note some of the important characteristics of growth velocity. The pubertal growth spurt is visible towards the right as a peak in velocity with a maximum of from five to ten centimeters per year. The timing of this peak varies a lot from girl to girl, ranging in this plot from ten to fourteen years of age. We also see at least one earlier smaller peak for most of these curves. But of course the highest velocities are achieved between one and two years, where they range up to an impressive 25 centimeters per year.

Now we can also look at the acceleration curves with Matlab command plot(hgtffd, int2Lfd(2)). Although we have to admit that the early accelerations are pretty wild, we can still see a lot of consistent features in the curves from six years of age on. The pubertal growth spurt is visible as a peak followed by a valley, and likewise for the earlier spurts.


Figure 4: The height accelerations of the first 10 girls in the Berkeley Growth
Study after penalizing roughness at the level of the fourth derivative.

Previous - 1 - 2 - 3 - Next