Monday, February 09, 2009

Relationship between Months of Inventory and Price Changes in GVREB

Here is an update on the refined half-over-half price changes to months of inventory (3 month moving average) relationship mohican and I optimised last year.
{
Nerd_verbosity++;

First the residual plots.



And the 95% CI fitted line.



We can see even with the deviation of last month we have high degree of confidence. It is easy to show a frequency component of seasonality so a high correlation is somewhat expected. Still, there ain't no purdy red dots on the bottom left and top right of the graph so it tells me a storm's a-brewin'.

And for completeness, the time plot.



Nerd_verbosity--;
}

Now, that was 30 seconds of your life you will never get back. Carry on.

6 comments:

puhlaya said...

Ohhhh, the residuals plot looks like it was created in MATLAB. Subplot function?

I'm more partial to the subplot1 function myself - it allows you to control the spacing between the individual plots (which is more useful for me).

wow. I spend too much time in MATLAB I think.

:-)

mohican said...

ubergeek!

awesome jesse

Definitely more pain to come.

Virgil said...

Looking at the last graph, does anyone have any thoughts on why there is a divergence between 'predicted' and 'actual' '% change HOH' for Dec 2008?

Warren said...

I think overall volume is so low in the winter months that it can skew the stats a little.

It looks like your prediction is a 1-month leading indicator in the spring-summer though, almost perfectly.

jesse said...

Overall data in the winter can skew but that is part of the reason for using 3 month moving average. There is generally more price weakness in the fall and winter when months of inventory is higher. My more detailed thoughts on this are available by following the link in the post.

Inventory Management Software said...
This comment has been removed by a blog administrator.