The problems were better highlighted in a radio programme that I can't immediately point you at, but were partly that in the attempt to get information specifically from Muslims, their very small sampling efforts were concentrating upon residents of areas in which Muslims were concentrated in virtually ghettoised conditions, emphasising any lack of social assimilation and sidelining those that do not consider themselves tied to the traditionalist side of their culture.
Like how areas with low 'immigrant' populations are far more anti-immigrant than those (though similarly British WASPs, themselves) that live and coexist with a thriving multicultural community around them. The London borough of Havering (randomly chosen from likely places, for the purposes of this post, not having time to scour for "the best example") seems to have around 85% explicitly white British Isles population, "other white" at 3%,whilst black and Indian populations, together, add up to around 6%, with the difference in small mish-mashes are in sub-percentile quantities. Lambeth (from the same source, chosen similarly cadually) is 40+% WBI, 15ish% "other white", 25+% B&I, and the larger remainder the mish-mash of various others. And when you then look at not only the total polling of each area, but even specifically the same major demographic of British/Irish Whites, they exhibit totally different opinion spectra...
The trouble is that it's easier to go "we need to talk to a Muslim, try this (Bradford/whatever) phone number and see if the person who answers is one of them!" to get your 1000 person 'sample'
1 than to dial nationwide (within
and outside the concentrated populations, proportionately) and get so many misses as you find that the farmhouse in Auchtermuchty doesn't have anyone you want to talk to in it, nor that holiday cottage in St Ives, or the places in Llangollen, Matlock, Edmundbyers, Keswick, Brigg, Goole, Bath, Malvern, etc, etc, etc (you're not allowed to ask if they know of any
neighbours that you can call who are Islamic...).
Thus you aim at area codes (and sub-area grouped phone pre-fixes) that correspond to an area that is positively culturally islamic, and get (in your 0.03ish% of your 'target' supersample) a lean towards an atypical viewpoint.
And that's
before you cherry-pick. You ask them "do you like this, do you not like that, do you have strong or weak opinions about the other", and many other questions. Having already briefly talked about the p-value (down there in the footnote), the more likely you'll find a question answered in a way you'd "like" to report, while ignoring the rest. Say that perhaps 97% of Muslims do
not support suicide bombing (do 3%, worryingly, support it, or is that 2.5% "I don't want to answer that", like they did
every question since they got bored, and 0.5% more like "hey, if it's against someone who is trying to kill my family, of course..."? The details matter...), but that's not a good figure to release. Go for the blip! Forget the neutral (or anti-blip) results! Sensationalise!
So take such weak/deliberate studies with a pinch of salt. (Not that I can work out how to do a better one, but it's likely worse than useless.)
But, if you want the Grauniad a second time:
https://www.theguardian.com/commentisfree/2016/apr/15/channel-4-islamophobic-bandwagon-british-muslimsDeep behind this story, there exist those lies, those damned lies, but mostly those very selectively used (and obtained!) statistics.
1 Out of about 2.7 million actual British Muslims... What's the p-value for that? Yeah, that's another point of contention.