# Common statistical tests are linear models (or: how to teach stats)

Thanks to my colleague Jenny Terry for tweeting this. More information at https://lindeloev.github.io/tests-as-linear/

# Worked examples for Gorard (GS) and Allen and Vignoles (D) measures of segregation in schools.

I put together this paper of worked examples for the Gorard (GS) and Allen & Vignoles (D) segregation indices. It measures the extent of school segregation within a geographical area using Free School Meals (FSM) as an indicator.

# Freaky teaching: Teaching Bayesian Statistics to Undergraduate Students using ventriloquist dummies

By virtue of its title Innovations in Education and Teaching International sets its aspirations very high. I’ve just been browsing the latest printed edition where Stewart and Stewart’s article ‘Teaching Bayesian Statistics to Undergraduate Students through Debates’ grabbed my attention. Like most people I have been educated in classical (or frequentist) statistics and have virtually no knowledge of Bayesian statistics—hence the title caught my attention. The article has therefore reminded me to rectify my ignorance.

The title gave me no clue to what I was about to read about though. The lecturer (second author Wayne Stewart) performs debates between ‘frequentists’ and ‘Bayesians’ using two ventriloquist dummies—Freaky the frequentist and his opponent, a doll depicting Thomas Bayes (c.1701-1761) in his persona as a Presbyterian minister (photographs appear in the article).

Personally I find ventriloquist dolls pretty sinister. However as an approach worthy of the label ‘innovation’, it’s going to be hard to surpass. Most of the students found the dummies funny though and claimed to have learned from them.

Sepideh Stewart and Wayne Stewart (2014). “Teaching Bayesian Statistics to Undergraduate Students through Debates.” Innovations in Education and Teaching International 51, no. 6: 653–63.

# Statistics for Humanities now in PDF format plus some reflections

The Statistics for Humanities book is now available in a much nicer PDF format. The mediawiki built website is still up there though it doesn't look great. The PDF is not perfect by any means but it’s time to move onto other projects now.

The book is published under a non-commercial Sharealike Creative Commons licence. I am thinking of trying an e-book version (I have software which technically supports this)  but I fear that with all the images and mathematics this could end up being a big mess.

The first versions of the book (previously available for public consultation and rejected by the British Academy on the advice of their reviewers) were written in LaTeX. I actually learnt LaTeX for the specific purpose of producing the book. I liked the idea of programmatic control but it all got a bit difficult to control where graphs and images landed. GNU plot which integrates with LaTeX was great for making the graphs and I'll continue to use it in the future if and when the need arises.

The first website version has been written using mediawiki software (the same software behind wikipedia). Wiki markup is easy to learn and plugins mean that LaTeX can be integrated for the mathematical parts. One of the issues with websites is that appearance in browsers can vary, but it is still not as nice looking as wikipedia.  The great advantage of wikis of course is that they are social tools in which people can collaborate to produce a finished product. I floated the idea with a couple of others of making this an open wiki, but was advised I could face problems of vandalism and spam. Although I like running websites it’s not a big part of my 'day job' and I don't want a troublesome hobby.

As I mentioned before the PDF version is available under a creative commons licence which allows people to modify it. This is all very well, but modifying a pdf document is not easy (that is sometimes the actual point), so I do need to think about whether I want to make the source files available. The book was produced using a combination of Serif Page Plus X7 (the full version) and MathType V6.9.

PagePlusX7 (I paid just under £64 including VAT) is not unlike MS Publisher though I much prefer it. I experimented with other desktop publishing software such as Scribus (free/ open source)*, but felt the learning curve was too steep at the present time. (High level DTPs run into hundreds of pounds so I'm not even going there).

Mathtype 6.9 (£43.20) was a worthwhile investment. I used it to type equations in LaTeX then copy and paste into PagePlus, but it can also be used to 'build' equations like the MS Equation Editor and even has a 'handwriting' feature. It can also be used in hundreds of other software applications. This was not totally plain sailing and for some reason equations using square root signs look a bit strange when posted into PagePlus. (If future editions of PagePlus have LaTeX Math integration that would be delightful).

Despite all the software used to produce the book, no software is necessary to use the book!

*Scribus does have LaTeX integration.

# Taxpayers' cash should not be used to fund faith schools, say voters (But what did the survey really ask?)

Taxpayers' cash should not be used to fund faith schools, say voters. Labour wants talks on teaching of religion as poll shows 58% of the public urge abolition or axing of state funds

The above headline in The Guardian intrigued me so I thought I would take a look at the actual questions asked. The survey was carried out by a company called Opinium on behalf of The Observer. On Twitter I tried to find out what questions were actually asked; interesting Richard Adams, Education Editor at the Guardian did not know, but said he would do some digging. I eventually found a pdf of the survey results on Opinium’s website.

In the interests of full disclosure I am a Christian with two children, one of whom is pre-school and the other attends a non-faith primary school. I did not attend a faith school myself. I’m not that interested in faith schools per se, but headlines such as the above invariably lead to debates about the relationships between religion and society that I feel duty bound to take an interest in.

The reason I’ve written this post is that I am very interested in questionnaire design. Questionnaire design is harder than most people think, but one would have thought a company which does surveys in its day to day work would be quite good at designing questionnaires. This one is so lousy I can only think it was politically motivated. It fact it is so bad I’m not entirely clear what the political motivation might have been.

Let’s start from the very beginning and take a look at Question 1. Respondents are asked to choose the opinion which is most like their own—note carefully -- not their opinion but the one that “best describes” their view of faith schools.
There are only three possible opinions plus a ‘Don’t know/ No opinion’ option.

The three available opinions are:

1. I have no objection to faith schools existing and being funded by the state
2. I have no objection to faith schools existing but they should not be funded by the state (i.e. private schools may be faith schools but not state schools)
3. Faith schools should be banned entirely

The best describes bit might be some sort of attempt to meaningfully take into account some of the nuances involved in such arguments but how views such as the following, which I suspect are widespread, fit into these three options.

1. I have no objection to faith schools existing and being funded by the state as long as they are only Christian/ CofE/ Catholic/ Muslim etc. (delete as appropriate)
2. I have no objection to faith schools existing and being funded by the state as long as they are NOT Christian/ CofE/ Catholic/ Muslim etc. (delete as appropriate)
3. I agree with faith schools but don’t agree with private faith schools because I don’t agree that there should be private schools.
4. I agree with faith schools as long as they don’t take their faith aspect too seriously.
5. I think all schools should be faith schools.
6. I agree with taxpayer funded faith schools as long as it’s not a really weird religion like…

As 1 and 2 above indicate there is really no such thing as a faith school, merely schools which are in one way or another connected to a specific faith, religion or belief system. Faith schools vary of course within faiths —Oasis academies, CoE voluntary aided village primary schools and ‘fundamentalist’ Christian private schools could all be described as Christian faith schools but they are certainly not the same in their ethos.

Question 5: In your view, how serious a risk is there of some predominantly Muslim schools encouraging their pupils to adopt extremist views?

1.Very serious
2.Quite serious
3.Not very serious
4.Not at all serious
5.Don’t know/ no opinion

The only specific religion mentioned in the survey is Islam. I’m not sure how one could answer this question in any meaningful way. This is in the context of the recent ‘Trojan Horse’ controversy of course, the full facts of which have not really emerged. The ‘Trojan Horse’ schools are not faith schools, or Muslim schools, but schools in which most children (and their parents) identify as Muslims. Overall 74% of those questioned thought there was a serious risk (i.e. very serious or quite serious), but what do we mean by extremist views? Extremists, however defined do not need extremist schools to incubate their views and activities. A lot of teachers would probably feel they are being credited as having far more influence on young people than they actually have.

A couple of further points: Although the findings are available there is a statement of confidentiality in the footer of each page. This may have been an oversight, or it might be a case that this data was never meant to be made public. I’ll assume the former until I have evidence to the contrary. It is also unclear how the survey was carried out, though it is possible to register with Opinium as a 'survey filler in' and I suspect people who fill in surveys are less likely to be in the ‘No opinion’ camp.

Whatever the truth the 58% figure does not strike me as particularly likely or unlikely—I’m not even sure if this is higher or lower than what I suspect would be the percentage of people disagreeing with schools having a religious ethos or culture. The Observer is evidently trying to influence the Labour Party on the matter of faith schools. I just wish they had designed a better survey.

# My Open Educational Resources

I have put many of my teaching resources in the humbox where anyone is welcome to download and reuse them. They are primarily used as handouts in face-to-face contexts, so most would probably need adapting to turn them into materials suitable for purely online context. Most are available under creative commons licenses, but please check first.

My profile on the humbox

Click on "View all resources" to see a full list of my resources.

The PhD research training collection includes resources on the PhD viva, academic writing, employability, academic writing, doing book reviews and ethnography.

The Head of University language departments collection includes scenario planning exercises, for curriculum change, people management and presenting research.

Further resources I have produced for sharing are on other project websites such as SPEAQ (Relating to quality assurance and enhancement) and Getting the Most Out of Feedback which has resources for students and staff on feedback and student evaluation.

There is also my Statistics for Humanities online book.

I have made a list on Diigo of other people's open educational resources on study skills

# Neave's statistical tables

I had an unexpected surprise when I 'won' some Taylor and Francis vouchers on Twitter. One of the books I chose was Henry R Neave's Statistics Tables: For Mathematicians, Engineers, Economists and the Behavioural and Management Science.  In the preface to this second edition the compiler writes:

[Teaching statistics In the 1970s] I was unable to find any tables suitable for the course. So there really was no alternative but to develop a set myself (p.3).

One of the challenges of writing Statistics for Humanities was actually finding statistical tables which did not have a copyright on them or which were otherwise non-ambiguous in terms of intellectual property. Requests for information on twitter and mailing lists of copyright experts were inconclusive-- some stating that the tables belong to the persons who complied them and other citing 'fair use' or the view that ideas or numbers cannot be copyrighted.

Either way I'm now in possession of Neave's tables. This doesn't clear up the issue in any way, but I'm pleased to have a copy to hand.

# Statistics for Humanities now on Twitter

I've now set up a twitter account for the Statistics for Humanities site.

# Why applications to study languages might not be declining as much we think.

Headlines such as Times Higher Education’s “Ucas stats reveal languages decline” have become an annual ritual in recent years. Group R (European Langs, Lit & related) down  from 22,486  to 21,248 applicants ( -1,238  -5.5%). Group T (Non-European Langs, Lit and related ) went from 6,678 students in  2012 to 6,241 in 2013 (-437  -6.5%).

I was struck by the vast increase (28.7%) in the number of applicants classified in Z. General, other, combined and unknown. In fact 297,071 students (11.9%) are classified either in a Y category (Combined arts, combined sciences, combined social sciences, sciences combined with social sciences, sciences combined with social sciences or arts, social sciences combined with arts) or in category Z.  Given the high proportion of language students who are doing combined/ joint degrees I suspect (but can’t prove) increasing number of linguists are finding their way in categories Y and Z.

2013 UCAS statistics