The roles of statistics in research are as follows:
1. Tool to calculate the amount of members of the sample taken from a population. Thus, the number of samples required is more accountable.
2. Tool to test the validity and reliability of the instrument. Before the instrument is used for research, it must be tested the validity and reliability first.
3. Techniques for presenting the data, so that data is more communicative. Data presentation techniques include: Tables, graphs, pie charts and pictograms.
4. Tool to analyze the data such as testing in hypotheses proposed. In this case, statistics used are correlation, regression, t-test, anova and others.
Various Kinds of Statistics:
In the narrow sense, statistics can be interpreted as data, but in broader sense statistics can be interpreted as a tool, tool to analyze and tool to make decisions. Statistics can be divided into two; descriptive statistics and inferential statistics.
Furthermore, inferential statistics can be divided into parametric and non-parametric statistics. Descriptive statistics is statistics used to describe or analyze the statistical results of the study, but it is not used to make broader conclusions (generalization / inference). Research that does not use samples, the analysis will use descriptive statistics. Similarly, research using the sample, but the researchers do not intend to make a conclusion on the population from which the sample is taken, the statistic used is descriptive statistics. In this case, correlation and regression techniques can also serve as descriptive statistics.
Inferential statistics is statistics used to analyze the sample data, and the results will be generalized (interference) for populations in which the sample is taken. There are two kinds of inferential statistics; parametric and non-parametric statistics. Parametric statistics is used to analyze interval data or ratio, which is taken from a normally distributed population, while non-parametric statistics is used to analyze nominal and ordinal data from a population-free distribution. So there might not be normal. In this case, correlation and regression techniques can serve as inferential statistics.
