Constituency by constituency threads for the next GE: utilising the P.ie hive mind



hollandia

Well-known member
Joined
Sep 11, 2012
Messages
30,147
SDs will have 2 minimum.
You would think so, but that's on the 1% attributed to them in this poll. It gets a bit fuzzy in the lower numbers unless the %ages have at least one decimals. 1% could equally be 1.49% rounded down.
 

hollandia

Well-known member
Joined
Sep 11, 2012
Messages
30,147
You can't identify 29% don't knows a week before Christmas 2018 as being out of the equation any more than you can claim the 71% will actually vote. It's really not good data. GIGO as they say!
Aren't all polls like that?
 

sic transit

Well-known member
Joined
Jan 30, 2008
Messages
25,414
Aren't all polls like that?
The ones closest to polling day tend to have pretty small DK contingent and usually end up a lot closer to the result.
 

hollandia

Well-known member
Joined
Sep 11, 2012
Messages
30,147
The ones closest to polling day tend to have pretty small DK contingent and usually end up a lot closer to the result.
Yes, that's true, but they are a fraction of all polls. And we only comment and speculate on the ones in front of us.
 

locke

Well-known member
Joined
May 2, 2007
Messages
3,375
That poll in the Sunday Indo., crap only 920 people were surveyed with a margin of error +- 3.2%
It isn't a great sample size admittedly, but remember that the 3.2% only applies to a party getting 50% of the vote and that it declines as the numbers go down. It's still around 3% for FG and 2.5% for Sinn Féin, but it's 1.4% at Labour's level and 0.6% once you are down at the Green and Social Democrat level, where rounding is more important than Margin of Error.
 

hollandia

Well-known member
Joined
Sep 11, 2012
Messages
30,147
That poll in the Sunday Indo., crap only 920 people were surveyed with a margin of error +- 3.2%
There's no poling company polls more than 1200. 1000 is typically the norm. The single exception to that was RTE's Presidential Election exit poll with almost 3500 respondents. a 3% MOE is also pretty much the norm.

Next Irish general election - Wikipedia
 

hollandia

Well-known member
Joined
Sep 11, 2012
Messages
30,147
Poll percentages can not be directly linked to actual seat numbers. V flawed method
No, they can't. But they can be analysed on a constituency by constituency basis, with weighted biases per previous party support (and inter constituency party split) to give a decent estimate of how things would pan out, assuming all other things remain equal.
 

Breanainn

Well-known member
Joined
Aug 23, 2014
Messages
2,950
You would think so, but that's on the 1% attributed to them in this poll. It gets a bit fuzzy in the lower numbers unless the %ages have at least one decimals. 1% could equally be 1.49% rounded down.
Gavan Reilly tweeted that the poll gave them 2%?
 

Robyn30

Member
Joined
Nov 26, 2017
Messages
57
No, they can't. But they can be analysed on a constituency by constituency basis, with weighted biases per previous party support (and inter constituency party split) to give a decent estimate of how things would pan out, assuming all other things remain equal.
this is true. But that's not what was done a couple of pages back
 

hollandia

Well-known member
Joined
Sep 11, 2012
Messages
30,147
this is true. But that's not what was done a couple of pages back
Which post are you referring to? Because if it's mine, that's precisely what I have done.

(Via what is technically termed a big feckoff spreadsheet)
 

hollandia

Well-known member
Joined
Sep 11, 2012
Messages
30,147
Gavan Reilly tweeted that the poll gave them 2%?
Yep, you're right - 2%, but the point stands - it could equally be 2.49% rounded. This can be the difference in being in the running or not.
 

Socratus O' Pericles

Well-known member
Joined
Oct 12, 2009
Messages
32,972
This thread is carrying on elsewhere...just saying.😀
 


New Threads

Most Replies

Top Bottom