VS

A note to everyone who was bashing away at each other on the Josh 14 thread, VS has added a comment to the end which may be of interest.

Comments

[Josh]

I am glad VS came over here - he likes the cartoons too, so much so he bought some t shirts and prints. (proceeds to go to good causes, of course).

[VS]

:)

[Lord BeaverBrook]

Josh
Love the unit roots can't wait for the visualisation of cointegration hope you've gained some insight!

[Richard Drake]

Let me atone for my misdeeds in this area by reproducing VS's comment, with working inline links this time, for your readers' convenience and delectation:

Hey, missed this one.

Everybody chill! :) [I'm trying, really I am :)]

Here's my 'trend' post, because there seems to be some confusion as to 'what I'm saying'. I think this post is perhaps the most relevant one I made. Links to all temperature related tests/simulation posts are in the second section.

Finally, since my exchange with Tamino is being mentioned, here's my reply to both his blog entries (with links to all relevant exchanges). Note that the first link contains links to monte carlo simulations (performed later), confirming claims made in this reply w.r.t. BIC lag selection in the ADF, and the PP test.

(Intelligent) review welcomed!

Cheers, VS

PS. For the record, different orders of integration - Temp I(1) and GHG's I(2) - prohibit *linear* dependence (i.e. regular cointegration). The variables can still be polynomially cointegrated (i.e. the level relationship can still exist, albeit in a funky form. To be fair, we would expect nothing less from our atmosphere :). This hypothesis can be statistically tested. BR do that.

Note that this is their work, not mine (they are the TSA experts, not me). I just happen to have read it, and found it very convincing. I want to see more of that stuff :)

[Don Pablo de la Sierra]

@ Bishop

I can hardly wait to see the cat fight when you finally post the Josh 16 (I hope) cartoon on the Presence of a Unit Root.

Perhaps VS will be good enough to referee :)

@ VS

So they could have used a polynomial regression, even though the assumptions of it are linear. I use to be the statistics consultant at Cornell University's computer center some forty years ago and I would have suggested that analysis (polynomial), but that was before DF and ADF tests were generally available.

And your are right -- the results of a polynomial would have been funky. But it would have given an idea how many factors were involved. Just what you do with that, I don't know.

I had some fun trying to explain Unit Root to Josh the other day and sorta kinda got it, I think, but I haven't looked at linear algebra for those 40 years and I didn't have a math book to refresh my memory. If he follows my suggestion and sends it on to you for your comment, please remember I was trying for a layman's level, not a mathematically rigorous treatment.

In any case, all of us without a strong background (including me) in linear algebra could use a layman's explanation of what the unit root is and why it is so important. Perhaps you know of a reference or could create it.