The Management of a Student Research Project by John A Sharp & John Peters

Author:John A Sharp & John Peters [Sharp, John A]
Language: eng
Format: azw3
ISBN: 9780566084904
Publisher: Taylor and Francis
Published: 2017-03-02T05:00:00+00:00

Quantitative techniques for explanation and prediction

Several techniques which are relevant to explanation and prediction will be examined briefly. As indicated in Figure 5.1 these are loglinear analysis, experimental design and regression.

The technique of loglinear analysis explains the variations in probabilities of class membership. For example, in the UK, the probability of a person being convicted of a crime before the age of 25 might be explained in terms of variables such as sex, socioeconomic status of the individual’s family, highest educational level attained, etc.

The experimental design model has, as well as its applicability in analysing many different types of research data, considerable virtues as a conceptual model of quantitative research directed towards explanation and prediction.

Fundamental to the model is the notion that the variable of interest can be measured on a ratio or interval scale and that the values of the variable to be explained or predicted are affected by a number of other variables usually referred to as factors. Each factor takes on more than one value and each value is called a factor level. Factors may often, however, be measured on nominal scales, for example, fertiliser A, fertiliser B or represent fairly crude groupings – for example, application of less than 100 grams of fertiliser per square metre, application of more than 100 grams per square metre, and so on. Finally, we assume that for each combination of factor levels we have at least one measurement of the variable whose value is to be explained or predicted. The model assumes this value is made up of a number of components: a base value plus various additive effects due to each of the factor levels and also due to interactions between factor levels, plus finally a random term representing errors in measurements, the effects of factors not considered directly, and so on.

The virtues of the experimental design model as a conceptual model of the processes involved in explanation and prediction are very considerable even if, for whatever reason, no attempt is made to carry out a statistical analysis. It offers an explanatory framework which is capable of handling complex relationships between the respondent variables and factor levels along with predictions of the effect of any particular set of factor levels. The factor levels can be recognised as independent variables and the implicit requirement of the model that there be at least two levels of each factor enables their effects to be isolated. The experimental design model assumes that the researcher can control the experiment to the extent of selecting the factors and factor levels whose effects are to be examined. This is, of course, not always the case in the social sciences but it may still be possible to approximate to an experimental design by using the fact that particular variables vary between one organisation or country and another or over time. Such applications of the model are usually referred to as ‘quasi-experiments’ (Cook and Campbell, 1979). The idea of quasi-experiments was originally developed in the context of education

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.