Analysis 103
haste, and overconfidence—and a lack of willingness to listen to criticism—leading
to difficulties later on.
Analysis
Two key concepts in every aspect of managing data and presenting results, which
I have touched on a couple of times in this chapter, are variables (or parameters)
and category. These concepts reflect our understanding of the data. We want to
understand what kind of data we have—what sort of ‘meal episode’, for example.
Assigning instances to categories lets us discuss and analyze data in a consolidated
way. Variables determine the behaviour of the data, and we have understood what
is going on when we can accurately predict how variables and data values interact.
These concepts underpin how we proceed with data analysis.
Having presented the data in an informative way, how much further should the
analysis go? In the work discussed earlier, Geoff made some general remarks about
his observations before he tabulated the results. The chapter stopped abruptly at the
end of the tables. He felt he had a good reason for doing this but, as I look back
on the work, I find it unsatisfying. He had some strong hypotheses to test. He had
designed experiments to test them, and had carried out the experiments. Were the
hypotheses upheld or rejected? The reader wants to know what the findings are
before the writer goes on to discuss their implications.
And it is not only the reader who is learning. Your presentation of results is part
of your process of interpreting them—writing the results chapter is part of a cycle
of understanding, not an end point. Your aim is to educate others, but self-learning
is likely to be part of the process, even at this late stage of thesis writing.
In complex situations such as the coal example above, in which there is con-
siderable interplay of the effects of the different variables, it may not be easy to
disentangle the results and their implications. Nevertheless, I recommend that you
try—plot the results in terms of the hypotheses. Geoff could have plotted ignition
times (a dependent variable) against the moisture content of the particles (an inde-
pendent variable), with all other independent variables held constant, or against the
size of the particles, or the oxygen content of the gas, or its temperature. He did
these plots for his own information; but he decided not to present them as part of
the results chapter because he had found that the interpretation of the results was
not as straightforward as his original hypotheses indicated. In retrospect I believe
the reader would have been in a better position to go on to the discussion chapter
had he presented the results as tests of his hypotheses by plotting dependent vari-
ables versus independent variables. Then, in a very brief discussion, he could have
pointed out the unexpected complexity, and announced that he would be dealing
with this in his next chapter.
But don’t go to the opposite extreme. In one thesis I examined the candidate had
discovered the power of a chart-drawing facility. His results chapter contained over
a hundred charts plotted by trawling through all of his data sets and plotting every