Different authors use different functions differently

[Chris Butler]

Dear Yuri,

Statistical packages are indeed sometimes quite difficult for linguists and
other researchers who are not statisticians. But they are extremely useful -
sometimes essential - and it is worth getting to know how to use at least
one of them, for anyone whose work is primarily quantitative. One of the
most important aspects of using statistics in linguistic research is knowing
the basis for the various measures and tests that are available, and what
criteria you can use in order to decide which measure or test is appropriate
for your data. This is something you need to know, whether you are using a
statistical package such as SPSS or doing the calculations manually.

However, although it is not too hard to do the calculations manually for
simple tests such as chi square or the t-test, it is much harder for more
advanced tests, and almost impossible for some techniques such as various
kinds of multivariate analysis on large data sets, which require the use of
computing power. Manual analysis is error-prone, and this is avoided by
using a computer package. On the other hand, the statistical packages often
give you a lot of output, some of it very important, other parts of it less
so, so that you have to know how to interpret the output.

There are some good books around to help researchers to get to grips with
the standard packages. For SPSS I would recommend the 3rd edition of Andy
Field's 'Discovering Statistics Using SPSS' (Sage Publications, 2009), which
is comprehensive and detailed, but well explained and even entertaining.
There is also the statistical programming language R, which is being used by
an increasing number of adventurous quantitative linguists. For this, there
are at least two recent books: Stefan Gries' 'Statistics for Linguists with
R: A Practical Introduction' (Routledge [Taylor and Francis Group], 2009),
and Harald Baayen's 'Analyzing Linguistic Data. A Practical Introduction to
Statistics Using R' (Cambridge University Press 2008), which is rather
difficult in parts, but useful nevertheless. R can be downloaded freely from
http://www.r-project.org/, and free packages with some of the functionality
of the standard ones can be found (e.g. PSPP from
http://www.gnu.org/software/pspp/, which I have not tried myself), and may
be useful for researchers who do not have access, through their
institutions, to the (expensive) standard packages.

There are also books which are introductions to statistics for linguists,
but are not linked to specific packages. If you are interested in details of
these, please get in touch with me directly.

Best wishes,

Chris Butler

--------------------------------------------------
From: "Yuri Tambovtsev" <yutamb at mail.ru>
Sent: Wednesday, September 29, 2010 9:10 PM
To: <funknet at mailman.rice.edu>
Subject: [FUNKNET] Different authors use different functions differently

> Dear Funknet colleagues, if corpora of English texts is not one unity but
> a conglomerate, how can we use statistic criteria? I started the
> discussion what tools to use for the analysis because many of linguists do
> not use all these complex statistical packets correctly. The other thing
> is who knows what and how the data are being analysed in them. We have a
> sort of a "black box" which has the entrance in which you put your data
> and the outcome where you receive your results. You must be quite sure
> that the data are analysed correctly. The more simple criteria you use,
> the better. This is why I stopped using all the stat. packets and began
> using very simple criteria like the coefficient of variation, the
> Chi-square and the t-test. At least, I am sure about the outcome. More
> often than not, it is advisable to try you data manually than to use the
> statpack. Looking forward to hearing from you if you agree with me that
> using statpacks may give you strange results. The fact that so many
> scholar spoke on the list about how to use the statpacks showed that it is
> a burning question for linguists and other researchers who are not
> specialists in mathematics. Yours sincerely Yuri Tambovtsev,
> yutamb at mail.ru