I'm no statistician

In fact, statistics is one substantial reason I'm not Atta J. Turk, Ph.D.

However, even I think that Newsweek really needs to figure out what the fuck they are doing in their polls.

In their June poll Obama was up by 15, now he's up by an insignificant 3.

Alarming?

Well, not really when you look at the sampling.

June:

231 Republicans sampled, 23%, 36% with "lean Republican" included
324 Democrats sampled, 38%, 55% with "lean Democratic" included
307 Independents sampled, 35%

July:

315 Republicans sampled, 28%, 42% with leaners
324 Democrats sampled, 35%, 47% with leaners
357 Independents sampled, 33%

Imagine that, when you poll the same number of Democrats, but 84 more Repoublicans (increasing their number by about 25%) and 50 more Independents, McBombBomb does better.

I think the first Newsweek Poll was pretty much overstating bullshit; and the second is almost a mirror image of bullshit.

I think PEW or Gallup or some other period polls are far more reliable then this wacky and unweighted variance.

Polls should be much more forthcoming about their sampling.