12.07.2015 Views

Download - HSRC Press

Download - HSRC Press

Download - HSRC Press

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

multivariate techniques are much more readily available than was the case afew decades ago. The major threats to validity in the application of thesetechniques are clearly systematized in the majority of standard methodologytexts. Examples relate to the fact that certain types of techniques are applicableonly to certain types of data (compare the requirements relating to the level ofmeasurement — nominal, ordinal, interval, and ratio), or that certaintechniques may only be used provided the distribution of the data is normal, orprovided the sample has been randomly selected. Statistical considerations ofthis nature are, however, beyond the scope of this book.The most important reason for not discussing analysis as a separate theme inthis chapter is, however, to be found in the fact that the criterion of inferentialvalidity applies equally to data analysis. Analyzing data involves makinginferences. Even the relatively simple process of assigning data to certaincategories involves an inferential process. Although this inferential process isnot nearly as complicated as that which is involved in the interpretation ofresults, it would be an error to regard it as merely a mechanical process.Philosophers of science such as Winch (1958) have emphasized that theprocess of identification (for example, when a researcher decides that a givenform of behaviour belongs to category A rather than category B) contains ademonstrable interpretative element. The methodological requirement ofinferential validity, therefore, not only applies to the final stages of theinterpretation of results, but also to analysis and, as we shall argue in thefollowing sections, to every other form of decision making in the researchprocess.Free download from www.hsrcpress.ac.zaRESEARCH AS A LOGICAL PROCESS: THE REQUIREMENT OFINFERENTIAL VALIDITYIn this section we shall focus on the research process as a typical example oflogical argumentation. Irrespective of the exact nature of a research project, itis possible to regard any research project as an extended logical argument.Logical argumentationLarry Wright (1982: 4) defines an argument as the (usually) dispassionatemarshalling of support for some statement (or viewpoint, or conclusion orposition). In the same manner that an individual would argue a case ineveryday communication by citing evidence in its support, the social scientistattempts to muster scientific evidence in support of a specific point of view.More specifically, one may say that it is the aim of the social scientist todemonstrate the validity or invalidity of a given theory or model (or eveninterpretation, or finding) by mustering sufficient evidence.Apart from the considerations of validity discussed in the preceding chapters,we now turn to a further criterion of validity that is referred to as inferential106

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!