Big Eck's SNP bigs up a Facebook unscientific poll - Statistics Not Partisanship

I just saw this headline on the SNP website

Facebook poll shows majority for independence

For a start we all know the 'accuracy' of Faebook polling, indeed I was surprised that I hadn't picked up on it before now. It turns out it is one of those questions that get asked on facebook from time to time

Even the SNP report claims that 45% of such a poll is a majority with 42% failing. Now that places the result within any statistical margin of error. But not a clearance of 50% within a margin of error, especially as a Facebook poll cannot be weighted to take out any irregularities or non-representation of the sample group.

When I checked the figures for first preference as this is not a two question option as Caron has been blogging about, were:

Yes to full independence for Scotland 7447 44.3% of decided 46.4%
Yes to devolution max 1356 8.1% 8.4%
No 7239 43.1% 45.1%
Undecided 761 4.5%

Hardly a majority by any measure and too close to call by most I say. Obviously the SNP don't stand for Statistics Not Partisanship.

Lies, Damn Lies and Tory Crime Statistics*

Working as I do with data day in and day out the one thing you learn is that you have to compare like with like to get an actual comparison. There are little comments all over our spreadsheets which say when something started, moved or a calculation was changed. It not simply for us to keep a track on changes but also so that we know if data from different periods is comparable simply from the raw figures.

In 2002 what happened with violent crime statistics was that instead of the incident being recorded as violent at the discretion of the desk sergeant police were told to record every complaint.

This is what Chris Grayling the Shadow Home Secretary failed to tell us when he was whipping up a storm with his some British town's as becoming like The Wire.

Sir Michael Scholar, chairman of the UK Statistics Authority sets the wayward Tory right by writing to him saying:

"I do not wish to become involved in political controversy but I must take issue with what you said about violent crime statistics, which seems to me likely to damage public trust in official statistics."

He included a note from his agency which said:

"It would not be appropriate for the Authority to seek to intervene in political debate directly. However, where we see that official statistics have been presented or quoted in a way that seems likely to mislead the public, we will publicly draw this to the attention of those involved."

Here are the actual figures, there is a kink up in 2002-3 and then the downward trend continued under the new recording criteria.

Grayling has said he would take the Statistics Authority memo into account, before then dismissing it totally by saying he would:

"continue to use recorded crime statistics, because they reflect an important reality; that the number of violent crimes reported to police stations, and particularly serious violent crimes, has increased substantially over the past decade, even taking into account any changes to data collection".

* There was just no way to top The Independent's headline.

A Question of Sexuality - or Not

As a family genealogy geek I've spent a lot of time recently looking through the PDF files of the Irish Census from 1911. Indeed some friends will doubtless tell you of my excitement as the time for the County Londonderry returns to go online earlier in the year.


I know the importance in history of the accuracy of the information that is given. I also like the fact that some relatives have given extra information* not required, but which for history makes things easier to look into a verify. Therefore when I heard that the Government were considering adding a question about sexuality to the 2011 census I thought about whether this was a good idea or not.

There are two official takes on it, first there is the one of my former colleagues at the Office of National Statistics (ONS) who say that it won't provide a true reflection of the number of gay, lesbian and bi-sexual members of society. Then there is view of the Equality and Human Rights Commission (EHRC) who say it is vital to know how many gays, lesbians and bisexuals there are in Britain, where they live and what jobs they do so the progress of equality legislation can be measured.

I'd say that both are partially correct. The inclusion of the question isn't going to give an accurate reflection, indeed one of the most misrepresented groups would be the bi-sexual population. Anyone who spends anytime around the gay community will know that they are people out there who will say 'I'm not gay but...' the but is of course that they will happily engage in sexual acts with those of the same sex. There will be bisexuals out there who are in monogamous (or not**) 'heterosexual' relationships and not all of them will be totally open telling their partner that is the case, so to fill it in on a census form would definitely be out.

The census is there to gather an accurate picture of the population, the question of sexuality is not even one that every member of society is able to answer. Coming out is hard enough for many of us, to possibly feel obliged to do so, or lie about it on a census form when others around you are filling it in is an issue. In fact looking to see what would have happened in previous census returns for myself in 1991 I was denying my own sexuality, so would not have given an accurate return to that question. In 2001 I had just moved in with my ex-fiancée's parents, who fortunately did know my sexuality, but could have been embarrassing otherwise. That is just me, and while I don't wear it on my sleeve it is not something that is hidden from people around, people who will also appear on a census return.

What about people in residences of multiple occupancy? I notice the wording of the question appears to only cover straight, bi or gay. So what about the whole grey area of transgender/transsexual identities? Or what about asexuality or polyamorous relationships? Indeed it is a very complex issue and not one that can be passed of in one question and tick box. A spokesperson for the ONS said:

"A suite of questions would be necessary to collect data on the different dimensions of sexual orientation, including attraction, behaviour and identity."

Even then are we using the Kinsey scale for each of the three aspects? Indeed passed on the three question option my answers will have been changing over the last two census years and will be different again at the next one.

Yeah I know we already have the optional question about religion, but people are more prepared to talk about that, or simply write in Jedi. However, there is talk of rephrasing that question as a two-parter, the first asking "Do you identify with a religion?" then the yes respondents filling in which.

Obviously I'm not advocating 'don't ask, don't tell' as that hides more things behind the scenes than a level of openness will provide. However, I'm with the ONS on this one. The question cannot be worded in a way that will provide an accurate measurement, at this time. For starters it is not a complete question, second while there many out there are gay and proud, there are also many who are bi and shy.

From personal experience I would say that many straight and especially religious people may be opting out of answering the question. Indeed many religious groups are already saying it is an invasion of privacy. There is also as I pointed out liable to be a skew to the gay and straight responses and the bi-sexual (and others) segment of identity is going to be under-represented through a number of reasons. Or course that is personal anecdotal opinion and not quantitative (which some will argue is the point of the census question) but I think if there was a way to measure it (which I don't think there is) it would bear out in practise.

*Examples include cities rather than just County or other UK Country. Widow(er)s including length of marriage and children details when this was only requested of couples.

** Though I'll focus on the monogamous ones rather than those that also cheat, not a condition restricted to gay or bi but also present amongst the straight community, in case Jan Moir is reading this.