.comment-link {margin-left:.6em;}

Wednesday, May 05, 2010

The unbelieveable claims of newspapers

Fresh from trying to compare its own poll to a previous one by ITV Wales, despite the fact that they were done by different companies, had different methodology and sampled different numbers of people in different parts of Wales, the Western Mail publishes another unbelieveable claim this morning.

It is one of many in this election and nearly all of them have rested on an incomplete understanding of statistical analysis. In other words opinion polls are a dangerous tool in the hands of the media, who use them to make the most bizarre assertions and often misinterpret them to fit their own message.

Once again today's paper tries to compare the latest ITV poll to one of their own, even though there is no statistical basis for such a comparison. However, they then go on to say:

The YouGov poll for ITV Wales shows a 7% swing from Labour to the Conservatives since the last election in 2005. If such a swing were replicated across the rest of Britain, the result would be right on the borderline between a hung parliament and an overall Tory victory.

Really? It is an attempt to pass off an orange as a pear.

The Western Mail are using a poll confined to Welsh constituencies to project outcomes across the rest of the UK. If that sort of comparision were sustainable then Labour would be heading for a majority government and Plaid Cymru would be set to make gains in the Home Counties. What absolute nonsense.

Lib Dems accuse someone else of misusing statistics. Come on Peter just a quick glance at any constituency literature from your party would prove that you are in no position to make these arguments.
Sorry, I would dispute that. The statistics we use are always valid. We do nothing different to any other party as I can evidence.
In some countries opinion polls are forbidden immediately (a week or so) before the election, because they can affect the voters. Maybe that's a good idea?
Peter, what about the statement on new Lib Dem literature that

"In many parts of Wales the contest is between the Lib Dems and Plaid, with the Tories and Labour not in the race."

This is factually incorrect.

There is only 1 constituency in Wales where that is the case.

But your leaflet has said 'many parts' so that it counts towards your national election costs and not your local (Ceredigion) costs.

That is an example of dishonest campaigning.
I have not seen that leaflet nor am I responsible for it. However it is not a case of misrepresenting statistics. I can also find literature from all the other parties containihg outright lies, distortions and misrepresentation. That does not make it right but it does not make it unique either.
"The statistics we use are always valid. We do nothing different to any other party as I can evidence."

Really! What about the bar graph that your Pontypridd candidate produced on his main leaflet which was based on only 614 people voting and although the Lib Dems won the seat it was with only 196 votes how deliberately misleading is that?
Again I have not seen this leaflet nor am I responsible for it but if it conveys the outcome of the election correctly and covers the trends then it is fine. It is certainly as accurate as bar charts I have seen from other parties (without exception)
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?