Do You Need Statistics to Understand Your Data?

As I think each of us in this group last week established, it is important to gain a strong statistical background. Most of us stressed that this was of particular importance within the study of psychology where new studies and theories are being researched and published regularly. Within these studies it is usual for raw data (eg. IQ scores, reaction times etc.) to be gathered. Now, the question is, what use is statistics with regard to this data?

I am going to address this question using two interpretations of the word statistics (because I am a ridiculously pedantic person who has even quibbled pedantically over the use of the word pedantic on more than one occasion). Last year we discussed the differences between qualitative research – which tends to produce narrative reports – and quantitative research which studies variables of varying quantities (remember me?). It is in the latter field that statistics is of vital importance for understanding the data that is produced. Statistics is not merely a mathematical subject, it is a tool for organising, summarising and simplifying data (descriptive statistics) and for analysing sample data in order to infer generalisations about the corresponding population (inferential statistics). No other tool is as useful as statistics for interpreting our findings, while also ensuring that said findings are both empirical and replicable.

However, an ability to use and apply the statistical method is not enough. Statistical outputs, as presented by SPSS, are indeed well organised and relatively concise; and with training and experience, one can read them easily enough. This, however, does not always result in researchers making correct and reasonable inferences about what their data reveals. Take the example of Herrnstein and Murray’s book, The Bell Curve, wherein it was claimed that inherited intelligence alone determined success and that social partition in America was caused by the separation, by IQ, of the “cognitive elite” and the “cognitive inferior”. According to them, the majority of the cognitively inferior were of African or Latin descent. How is it that statistically educated researchers came to this conclusion? Quite simply, they either misused statistical methods or they attempted to be too parsimonious. For, you see, they decided that success in life was determined solely by intelligence, neglecting the impact of environment and upbringing. Too much emphasis and importance was placed on one variable to the point of incorrect exclusion of potentially confounding variables. (Click here for James Heckman’s critique of The Bell Curve) In cases such as these, however, it is often further studies – where statistical analysis is used appropriately – that expose the weakness and problems of the previous research. Therefore, the moral of this tale is that while statistics are needed to interpret data, your statistical understanding of what the data is telling you may not always be correct. Statistics are essential for gaining a better understanding, but will only help you to do so, if used correctly.

To end on a less serious (and definitely nit-picky point), Gravetter and Wallnau taught us that while statistics is often used as shorthand to refer to statistical procedures, the word statistics on its own can refer to something else. A statistic is derived from a sample and is a descriptive characteristic of a sample. These statistics, if you remember, have a corresponding population parameter. More often than not, in behavioural and medical sciences, one cannot gain access to the full target population within a study. This is why we use samples as representative groups from the population. Without the characteristics/statistics of these samples to study, very few projects could be completed due to insufficient access to the population. Therefore, in the case of studies done using samples, statistics (as defined in this manner) are indeed vital in that they form our data. And then, how might we interpret and understand this data? Why, by using statistical procedures, of course!

If you bore with me through my fussiness there, I am extremely grateful 🙂

Speak to you all next week.