Students - Statistics - some myths debunked

STATS blues? See red at the mere thought of SPSS? Green with envy at the more statistically minded in your department? When it comes to stats, our thinking is coloured by associations with maths and a generally scary reputation. I can see a self-fulfilling prophecy in the making, so let's debunk a few myths.

You have to be good at maths to be good at stats
As Professor Christine Dancey, co-author of Statistics Without Maths for Psychology, told me, 'we have
to remember as psychologists that statistics are tools that we use, we don't need to learn statistics for its own sake'. Her book hardly has a formula in its 636 pages. What is more important than maths is the ability to relate statistical techniques to real-life examples.

Stats are OK, but SPSS overcomplicates it all
SPSS For Windows is there to simplify our life by carrying out all the calculations for us.
The 'Help' pages of the package are very informative, and yet few people bother to use them! Stats books have helpful annotated pictures of SPSS showing how to input data and how to read the output, and there are also books dedicated to the use of SPSS, notably SPSS for Psychologists by Nicola Brace et al. and Paul Kinnear and Colin Gray's SPSS 12 Made Simple. They take you through every single command, button and function you can think of, and they explain what those scary tables in the output really mean.
Speaking of scary tables, some students seem to think that the results section of a lab report is an exercise in 'copy object' and 'paste'. As Paul Kinnear notes, 'ANOVA tables can be particularly obscure without appropriate editing', and the way you edit them shows whether you can separate and retain only the important elements. Indeed, this is a good time to check your understanding of the procedure you are using. Of course, the temptation is there to keep, say, the Huynh-Feldt line, just in case it was what the lecturer talked about that day I bunked off/was surreptitiously listening to my MP3 player while hiding behind my big mate at the back of the class…

The fancier the technique, the better
This is the 'divination' approach to data collection and analysis and is particularly applied to final-year projects, when a student experiences the giddy freedom to develop their own design. Like a clairvoyant waiting for a picture to emerge from coffee grounds, the student expects something meaningful to arise from the trail of SPSS output. It's got to happen! After all the variables I put in, and all the fancy techniques I used.
Well, contrary to popular belief, less can be more. Why complicate one's life (and expose one's incompetence) by choosing a complex inferential test when a t test would do? Statistics are tools and the trick is to know why and how to use them. What are we hoping to find? What do we need to get there? There is no substitute for a clear design in your mind, bringing home again the importance of conceptual understanding (and of the saying, 'garbage in, garbage out').

Phew! So glad to see the back of the t test – Luckily we're not going to use it next year
Perhaps the most important principle of statistics is that you build brick by brick, in manageable stages (so that it is not too daunting). If the foundations have disappeared or were very shaky in the first place, no wonder that the edifice you are trying to build will not stand.
The importance of making links is reiterated by Andy Field, author of Discovering Statistics Using SPSS for Windows: '…continuity is really important. I tend to emphasise the similarities between different statistical procedures and as we move onto new topics I try to constantly refer back to previous lectures and so on. So, for example, my students (much to their horror, I'm sure) get reminded of what a model sum of squares is in pretty much every lecture for 13 weeks!'

Who cares about stats? I'm going into 'qualitative' research anyway.
In research practice quantitative and qualitative research methods complement and illuminate each other, and in many ways this dichotomy is artificial. If you would still like to tell statisticians what to do with their normal curve, it is a good idea to know what it is meant by 'normal curve' first! Sure, it is not merely a matter of methodology; the issues are further upstream at epistemological level. But – and I speak as a social constructionist – the best possible critique can be developed when there is an understanding of the assumptions and procedures used by the 'quantitative' camp.

Having debunked some common myths, let's hope that 'stats anxiety' as felt at Time 1 (before reading this short article) has substantially decreased (ideally close to 0) at Time 2! If you need more, see my interviews with stats books authors in PsychTalk this year.

- Toni Brennan is in the Department of Psychology, University of Surrey. E-mail: [email protected].