• Both comments and trackbacks are currently closed.
• Trackback URI: http://hominidviews.com/wp-trackback.php?p=1408
• Comments RSS 2.0

2 Responses to “Poll Analysis: Obama Gains Ground on McCain”

1. CoolAqua Says:
   

   Thursday, March 27th, 2008 at 10:42 am

   Darryl
   You’re doing a lot of work, and I commend your efforts. There are many different ways of making projections, and I don’t think anyone has a lock on the “correct” way.

   Because you update your results so frequently, they are somewhat reminiscent of listening to a stock market report on any individual day. A single days values just don’t provide that much meaning.

   With this in mind, I would recommend that you start to graph your results over time and present them along with your updates. That would put the data in more context by shhowing how it varies over time.

   While I beleive these projections are really a window into the past, I do beleive they are important because you can then start to juxtapose what happened historically a few days before the polls were updated, to see how the public is responding to various inputs.

   I am somewhat dubious on how polls and projections are made. I’m concerned that they don’t capture the disconnect between “favorability” ratings, and “who would you vote for” ratings. These rating are wildly diverging at the present time, this usually points to a problem when this happens in the stock market.

   Regardless of this, as long as you keep your methodology consistent, you data provides yet another view of this race.

   Please keep up the good work, and consider putting your results in some form of summary like a graph vs time.

2. Darryl Says:
   

   Thursday, March 27th, 2008 at 5:09 pm

   Hi Pete,

   “With this in mind, I would recommend that you start to graph your results over time and present them along with your updates. That would put the data in more context by shhowing how it varies over time.”

   I occasionally do that and typically post an animation of the map and electoral vote distribution at the same time.

   For example in this post on 17 Mar, I included this graph:

   

   The graph involved weekly simulations conducted over an eight month period. And, yes, each time I do one of these graphs, I redo all the simulations. Why? Because (1) some pollsters hold onto their polls for a couple of weeks before releasing them to the public, (2) sometimes I discover an error I made in an old poll, (3) sometimes I learn about old polls and add them to the data set.

   I could store all old results and then figure out what dates must be rerun, but it seems a lot easier to simply let the computer spend an hour re-running the old analyses.

   BTW: I’ll post a trend graph over eight months late this evening.

   “While I beleive these projections are really a window into the past, I do beleive they are important because you can then start to juxtapose what happened historically a few days before the polls were updated, to see how the public is responding to various inputs.”

   Yep…I agree completely. Plus…it is fun to know the “score” at any point in the game….

   “I am somewhat dubious on how polls and projections are made. I’m concerned that they don’t capture the disconnect between “favorability” ratings, and “who would you vote for” ratings. These rating are wildly diverging at the present time, this usually points to a problem when this happens in the stock market.”

   Maybe…I just take the polls as imperfect data and try to learn what I can from them. I do avoid projections and try to make clear that the results are what would happen if the election were held now. (BTW: I have done projections before, but I restrict that to a day or two prior to an election The projections involve different techniques.)

   “Regardless of this, as long as you keep your methodology consistent, you data provides yet another view of this race.”

   Exactly! It will be interesting to compare different methods and how they compared over time. I am particularly intrigued by the “markets.”

   “Please keep up the good work, and consider putting your results in some form of summary like a graph vs time.”

   Thanks. I’ll try to incorporate more time trend stuff perhaps by leaving out the animations, which are large and chew up bandwidth.