[tech-spec] VIX follow up

  • From: jared albert <jaredalbert@xxxxxxxxx>
  • To: tech-spec@xxxxxxxxxxxxx
  • Date: Thu, 23 Dec 2004 16:09:29 -0800 (PST)

I just finished my intro stats class and wanted to
revisit a VIX study I had posted a while ago. Needless
to write, after the intro course I know only enough to
run minitab but not enough to even hurt myself. 

In the last study, I had grouped percent changes in
the VIX from the previous day to the current day and
then I had looked at the magnitude of the percent
changes in the dow 5 days out and 30 days out. In the
last study, it seemed that the greater the magnitude
of the vix changes, the greater would be the change in
the dow over the next five days. Michael Cohn wrote me
that basically I had discovered that the noise was
normally distributed around the mean. At the time I
didn?t understand that at all. However today I
regressed the 1 day dow percent change onto the 1 day
vix percent change and got exactly that, a good t test
but an r-sq of ~0. I redid it using the absolute
values of both just to see if the magnitude was at all
predictive and got the following output:

 Predictor           Coef    SE Coef      T      P
Constant       0.0067199  0.0001812  37.08  0.000
absolutevalue   0.008981   0.001836   4.89  0.000

S = 0.00714839   R-Sq = 0.6%   R-Sq(adj) = 0.6%

Analysis of Variance

Source            DF         SS         MS      F     
Regression         1  0.0012228  0.0012228  23.93 
Residual Error  3666  0.1873307  0.0000511
Total           3667  0.1885535

The F is good. The SE is very low, but the R-sq is
also very low. I wish I could post the scatterplot of
%vix1daychange vs %dow1daychange. It?s a big circular
blob, normally distributed around 0. 

Anyway, although I know very little about this stuff,
I wanted to post a follow up.  

Other related posts:

  • » [tech-spec] VIX follow up