Statistic's: Part 1 of 3

Statistics

1. Introduction

In research, much data obtained will contain numerical values. Collecting data is first stage, but the second stage is being able to make sense of it. This is where the science of statistic’s come into play. Statistics is used and expressed in two ways. Firstly, in order to determine the averages and distribution with the data and measures of central tendency, this is known as ‘descriptive’. Alternatively, the second option, inferential, this is the need to analyse patterns and relationships within the data to enable testing of the research hypotheses (Eastman, 1984). 

The use of statistics helps to analyse patterns within the sample data and then to draw inferences about the wider population, as illustrated in Figure 1 below. The opted model of statistics will vary accordingly depending on the nature of query and there are three sets of conditions required. Subsequently, at the first point of call, a choice needs to be planned at an early stage. Secondly, the nature of the research question should drive the choice of statistical analysis. This aspects is determined by whether a researcher is looking for ‘association’s’ or ‘differential’. 

Figure 1: 

2. Correlation 

Correlation is used when attempting to determine whether if a relationship might exist between variables, denoting the overall strength. A correlation is a measure of covariance, which is the extent of 2 variables appearing to be linked with one another. A correlation provides an output; known as the correlation coefficient which denotes the strength of the relationship between the 2 variables (Walkers & McLean, 1973). 

If no relationship exist between the variables, the value of 0 (zero) is assigned or if there is absolute correlation then a value of 1 (one) is allocated. If there are more than 2 variables, a complex examination of the interrelationship of the variables would be required. The technique deployed towards this is Multiple Regression Analysis which can help answer the following 3 questions: 

• How well does a variable or set of variables predict outcome? 
• Which variable from a set is the strongest predictor of outcome? 
• Does a variable continue to predict outcome when others are controlled?

Multiple Regression (MR) seeks to examine one dependent variable and contrasted to a number independent variables. MR requires reasonable sound sample size of data set of which must be large in quantity otherwise there will be an issue with generability of testing the hypotheses.

3. Relation to Research

Figure 2:

3. Summary

Statistics are a fundamental part of quantitative research and should form part the research design process. Gearing statistics towards the research question aims to provide a more precise and testable answer.

References:
Eastman, B. D. (1984). Interpreting Mathematical Economics and Econometrics. Macmillan Education, London. 
Walker, J. & McLean, M. (1973). Ordinary Statistics. Second Edition. Edwarrd Arnold Publishers Ltd, London.