Friday, April 05, 2013

Are Mr. Cummins & Company Really Re-Rising?

ThereAreNoNumbersLike
CrunchedNumbersVille


Yesterday, I went back to the AReid tracker in an effort to support my hypothesis that the BC Conservatives may be set to begin rising upward once again.

Which led North Van's Grumps to note that, based on his number crunching, the Cons have actually been rising for a long, long time:



OK?


_______
And apropos of another recent post....Billy Bragg is in town for a show tonight...Christina Montgomery wants to hear...This!


.

7 comments:

Anonymous said...

be cool to have a graph that was interactive with the trend lines being adjusted via user selected start points - a slider along the botttom

ie: what would they look like if started when CC was selected v. when HST announced etc

Ed Seedhouse said...

Anyone can hang a line on a time series, but what's the mathematical justification?

What was used to generate the line? Ah, a check at their site says they used Excel. So what is the R^2? Offhand the relationship seems pretty weak to me.

My bias is to like the lines as drawn. Thus I am even more suspicious of the graph.

You do science. Is this really evidence? Or coincidence?



Anonymous said...

job #s out today are not good

US? in the tank too
http://economistsview.typepad.com/timduy/2013/04/pause-coming.html

RossK said...

Anon-At-The-Top--

Ya - Norm Farrell made that point about looking from just before the HST was announced forward

______

Ed--

NVG says he used trendline, presumably to generate a linear regression. True, he didn't add the correlation coefficients. Personally, I'm not that interested in linearity given the increasing swings/waves in the red and blue traces as you get closer to the present...However, what caught my eye was NVG's assimilation of numbers from more than one source over six years...Thing is, if you look where the Cons were during the first to years on the graph and compare it to where they have been for the last two years would you not agree that they are on a longterm upswing?

_____

Thanks Anon--

The Economist also had a good article on the overall Canadian situation recently. Personally, the case they make that consumer spending is our only avenue for growth actually suggests that a reverse consumer to corp tax shift like that which just happened BC might actually be a good thing, from a macroeconomic point of view.


.

Anonymous said...

Austerity was/is a disaster in Europe and it is killing their economy - I don't think China will be able to pull the world out of recession again (relying on communists to save capitalism was never a good idea).

Harper's killing the Canadian economy with his corporate & individual tax cuts (starve-the-beast).

Ed Seedhouse said...

First thanks for taking the time to reply. When I eyeball the Con curve it seems fairly apparent that virtually all of the upward slope is "explained" by the sudden jump in 2012. Take that away and the upward trend would, I think, have a much smaller slope.

The question that arises is, in any random series taken over the same pre-2012 period, how likely would such a result be to arise by pure chance. And unless P comes out at 0.05 or less I would hesitate to say we can be confident that the observed slope is real.

Of course the 2012 rise is easily explained by the arrival of Mr. Cummings as party leader and I have no trouble accepting that as real.

But take that away and it doesn't seem all that likely that the correlation coefficient would be sufficient to rule out the possibility of chance.

Of course this is all eyeball reckoning, and someone should do the actual math if they really want to know. Math trumps eyeballs.

Aimple correlation is, if I recall my study of stats correctly, not really the proper way to analyze a time series since one of the two variables is clearly not normally distributed. There are other ways of doing it which my study was not advanced enough to reveal unto me.

All Excel includes is correlation by least squares unless I am mistaken. I believe that is justified when both variables are normally distributed, but not for a time series, for which real tools are available in real math packages but not, unless I am mistaken, in Excel.

Just sayin'

macadavy said...

Nice graph, but we have the technology (!) to go where no one has gone before! Alas, Kim Rossmo was too good for the VPD but his Geo-profiling has gone global. Now we have another local genius who has one-upped him with this: http://thetyee.ca/MasterClass/Spring2013/Hugh-Stimson/
And I'll one-up your Billy Bragg (couldn't afford him this time, alas, last seen @ the WISE Hall many moons ago) with his re-lyricism of another classic: http://www.youtube.com/watch?v=Zk69e1Vcmvg
Complete with brass band & choir, as it should be!