I am calling Bullsh*t on the DailyKos poll....

[The Vorlon]

I am just about ready to call **BULLSH*T** on this poll...

Here is the real simple math...



NOTE - I fixed the 3 day numbers, I had Obama and McCain reversed on Sept 10 - this day is NOT relevant to the caluculations, but I did make anerror

McCain was up by 1% - (47 to 46%) on the three day average on the Sept 10-12th sample

Lets say McCain just got lucky on the rounding and the real result was

46.51 for McCain
46.49 for Obama

ie the race was really tied.

Today (The Sept 11-13th sample) Obama is up 3% (48/45)

Again taking the extreme rounding scenario lets say the real result is:

Obama 47.51
McCain 45.49

ie a real gap of 2% at absolute minimum

A FOUR point swing (+1 to -3) in a three day tracking poll requires the sample that drops off to be at the absolute worst case scenario of rounding to be 6% different than the sample that rolls on...

Yet the sample that rolled off was 48/45 in Favor of Obama, and the sample that rolled on was also 48/45 for Obama....??

At the absolute extremes of rounding, the gap between two different 48/45 samples could be 2%, yet to move the 3 day average a full 2+%, an absolute difference of 6% is needed...

I call bullsh**t on this poll...

Anybody wanna argue?

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

More Bulltweet..

The numbers don't even add up internally..

If you take the percentages of support by party id, and use the distribution given... they don't add up..

   GOP   Dems   Indys
Sample   0.283716284   0.384615385   0.331668332
Obama   6   83   42
McCain   90.51   12.51   46.51 (note I have rounded McCain down as far as possible)
         
         
McCain   GOP   25.67916084   
   Dem   4.811538462   
   Indy   15.42589411   
   Total   45.91659341   

Even rounding down as far as possible, McCain is still at 46%, not 45%....

Something does not add up here....

[minionofmidas]

The figures given in the thread have them tied in yesterday's 3-day, not McCain up one.

[The Vorlon]

The figures given in the thread have them tied in yesterday's 3-day, not McCain up one.

the results posted by DailyKos are as I have presented them.

here is a screenshot of todays DailyKos



Check the url yourself to verify if you like...

http://www.dailykos.com/dailypoll/2008/09/14

[Verily]

Maybe they actually had McCain up 48-45 in the older sample and just mistyped it.

[minionofmidas]

Well... check the thread, too.

I would suppose the numbers currently posted for previous days are wrong. Somebody misentering stuff in a spreadsheet.
At the very least, we'll have to consider it a possibility.

Maybe they actually had McCain up 48-45 in the older sample and just mistyped it.

Even then, the last three days' figures don't add up to that kind of lead. Several things fouled up here, I suppose.

[○∙◄☻¥tπ[╪AV┼cVê└]

You made a mistake. In the 3 day average listed for 9/10 (does that mean 9/8-9/10), Obama leads McCain 48-45.

[The Vorlon]

You made a mistake. In the 3 day average listed for 9/10 (does that mean 9/8-9/10), Obama leads McCain 48-45.

that is exactly what Ihave posted.. the three day was 48/45, the one day was also 48/45

there is no mistake.. please check and confirm..

The math just does not work....

even the 3 single day samples don't add up to the 3 day average

Obama's last 3 days are

46, 46, 48..

If we round up for all three days...

46.5 + 46.5 + 48.5 => 141.5/3 => 47.16 not 48....  this gives Obama the maximum possible benifit of the round...

Same with McCain...

His last 3 days...

46, 47, 45...

if we round down all three days the maximum possible..

45.51 + 46.51 + 44.51 => 136.53/3 => 45.51 which rounds up to 46......

This is the worst possible rounding scenario for McCain...

This thing does not add up...

[minionofmidas]

The figures that graph gives fo the 3-day sample aren't the ones reported in the thread - not for today, not for yesterday, not for the day before. They are , however, the same as the ones given in the graph for the 1-day sample. Shouldn't be hard to notice that.

You made a mistake. In the 3 day average listed for 9/10 (does that mean 9/8-9/10), Obama leads McCain 48-45.

that is exactly what Ihave posted..

Actually, no:

As to the bit you added, noted. Does indeed not add up (and he is reported at 45 in the thread, too.)

[Sam Spade]

Well, it's Daily Kos.  Caveat emptor.

[The Vorlon]

The figures that graph gives fo the 3-day sample aren't the ones reported in the thread - not for today, not for yesterday, not for the day before. They are , however, the same as the ones given in the graph for the 1-day sample. Shouldn't be hard to notice that.

You made a mistake. In the 3 day average listed for 9/10 (does that mean 9/8-9/10), Obama leads McCain 48-45.

that is exactly what Ihave posted..

Actually, no:

As to the bit you added, noted. Does indeed not add up (and he is reported at 45 in the thread, too.)

maybe I am missing something, but if I take screenshots as follows is see a 4 day trend in the 3 day average for Obama of 48/46/46/48.... just as I have posted...

I have not looked at the thread, only the actual numbers on the DailyKos website.. are the thread and the Kos website different?

[○∙◄☻¥tπ[╪AV┼cVê└]

I checked the numbers, and you do have McCain and Obama's 9/10 3-day tracking numbers switched.

Now, are we sure which days the 3-day tracking numbers are for? What if the 9/13 3-day tracking numbers were for 9/12-9/14, and 9/13 is listed just because it's the middle of that period? Now, I'm aware that the 9/14 single-day numbers haven't been released yet, but the poll on the main page does say 9/11-9/14. Yeah, it's weird that they'd have the 9/14 numbers released. There may be some sort of fencepost error involved.

Alterately, maybe 9/13 means 9/10-9/12, but I already checked, and that doesn't seem to work.

[minionofmidas]

Are you messing with our heads? Did you just correct the error in your own graph without admitting to it? (The last line there used to read 48-45-45-48-364. Now it reads 48-45-48-45-364, as it should. Just in case anyone reads this later.)

Yeah, the figures reported in the thread are 47-45 currently, 47-47 yesterday, I forget whether it's 46-47 or 47-46 the day before that (I think the latter.)

As I said, something clearly is very wrong here.

[The Vorlon]

The daily sample for Sept 10th is:

   Obama 48
   McCain 45

   http://www.dailykos.com/dailypoll/2008/09/11

The daily sample for Sept 11th is:

   Obama 46
   McCain 46

   http://www.dailykos.com/dailypoll/2008/09/12

The daily sample for Sept 12th is:

   Obama 46
   McCain 47

   http://www.dailykos.com/dailypoll/2008/09/13

The daily sample for Sept 13th is:

   Obama 48
   McCain 45

   http://www.dailykos.com/dailypoll/2008/09/14

Do we agree?

[The Vorlon]

Are you messing with our heads? Did you just correct the error in your own graph without admitting to it? (The last line there used to read 48-45-45-48-364. Now it reads 48-45-48-45-364, as it should. Just in case anyone reads this later.)

Yeah, the figures reported in the thread are 47-45 currently, 47-47 yesterday, I forget whether it's 46-47 or 47-46 the day before that (I think the latter.)

As I said, something clearly is very wrong here.

No.. No.. I am not playing with your head (actually I am - I did fix it) but the 4 day old 3 day average is actually ireleevant to the discussion.

Here is is in a nut shell.. let's simplify..

The September 10th sample (for one day) was 48/45 Obama.. correct?
The September 13th sample (for one day) was 48/45 Obama.. correct?

Taking the maximum possible rounding error, these two samples could thus be at most 2% different...  Correct..??

The 3 day sample moved 4 points overnight, from McCain 47 Obama 46 to Obama 48 McCain 45
Thus the sample that rolled off (September 10th) must have been at minumum, at the extreme of rounding 6+ % worse for Obama than the September 13th sample..?   Correct...?

But we know it was the also a 48/45 sample.. at max 2 points difference.... Correct?

I can't get this thing to add up, because it doesn't add up....

[minionofmidas]

No.. No.. I am not playing with your head (actually I am - I did fix it) but the 4 day old 3 day average is actually ireleevant to the discussion.

Indeed so.

You must be logged in to read this quote.

I'm actually taking your word for it.

You must be logged in to read this quote.

Correct.

You must be logged in to read this quote.

Incorrect.

The 3 day sample moved 2 points overnight. It was 47-45 on the 10th, 47-46 on the 11th, 47-47 on the 12th, and is again 47-45 on the the 13th.

Which, incidentally, doesn't really square all that well with the daily figures either.

[Rowan]

The last three single days are:

9/11- M 46%- O 46%
9/12 M 47%- O 46%
9/13 M 45%- O 48%

Thus the three day average for release today should be Obama 47%- McCain 46%.

Something is wrong with the poll.

[○∙◄☻¥tπ[╪AV┼cVê└]

The daily sample for Sept 10th is:

   Obama 48
   McCain 45

   http://www.dailykos.com/dailypoll/2008/09/11

The daily sample for Sept 11th is:

   Obama 46
   McCain 46

   http://www.dailykos.com/dailypoll/2008/09/12

The daily sample for Sept 12th is:

   Obama 46
   McCain 47

   http://www.dailykos.com/dailypoll/2008/09/13

The daily sample for Sept 13th is:

   Obama 48
   McCain 45

   http://www.dailykos.com/dailypoll/2008/09/14

Do we agree?

Yes. Barring some sort of transcription error on their end, those are the actual daily numbers. I noticed that you fixed the 9/10 3-day numbers, although those don't seem to be relevant to the question at hand.

Now, the question at hand is which 3-day periods do the 3-day ones correspond to? Barring some sort of transcription error, you ruled out it being the last day. If it's the day afterwards, both 9/09 and 9/12 have McCain up 47-46, so that's not it, either.


I think the two remaining reasonable possibilities are:
1. The 9/13 3-day numbers mean 9/12-9/14, and we don't have the 9/14 1-day numbers yet.
2. There is some sort of transcription error, most likely with the 9/12 3-day numbers.