Hw 4, #7
I got my Matlab program to give what I think are the correct bias and var for hypos a  c; at least a and c agreed with the book. I am not sure, however, if it ran correctly on d and e since there were large differences for both bias and var for different runs even at 10000 trials. It seems that these were very unstable. I'm wondering if there are roundoff errors.
Thanks. 
Re: Hw 4, #7
I'm seeing the same thing in my Matlab program but it makes sense to me. In both cases you're fitting a parabola to two points as an approximation of a sine wave (d uses a fixed vertex, e has a variable vertex). Think about what this means for g_bar of d and e? Given this effect on g_bar, what happens to bais and variance?

Re: Hw 4, #7
I ran the ax^2 model ([d] hypothesis) with N=2E6 points a few times and got a quite stable bias value (even with smaller N values) but indeed, as you mention, a "somewhat" wild variance value is seen in several runs, that is, it is not as "stable" as the bias.

All times are GMT 7. The time now is 10:27 PM. 
Powered by vBulletin® Version 3.8.3
Copyright ©2000  2021, Jelsoft Enterprises Ltd.
The contents of this forum are to be used ONLY by readers of the Learning From Data book by Yaser S. AbuMostafa, Malik MagdonIsmail, and HsuanTien Lin, and participants in the Learning From Data MOOC by Yaser S. AbuMostafa. No part of these contents is to be communicated or made accessible to ANY other person or entity.