Correlation analysis is a set of mathematically sound methods by which is detected correlation between a pair of factors or signs of having a random component.The set of methods used in this research methods are widely used:
- Construction of the correlation fields, drawing up correlation tables;
- calculation of the correlation relationship or random factors;
- testing the hypothesis statistically significant relationship.
Continued research leads to the establishment of specific types of relationships between variables.The relationship between random signs or factors which exceed three, needs to use the method of multivariate analysis.
field and table, which is engaged in the construction of the correlation analysis are used as aids in the analysis of sample data.Inflicting on the coordinate plane sampling points come to obtain the so-called correlation field.By the way the point is located, it is already possible to make a preliminary forecast and to determine the form of dependence of random variables.Numerical analysis of the results requires grouping them in the form of correlation tables.
first appeared in the XVIII century, the term "correlation" with a light hand paleontologist Georges Cuvier began to actively be used to shape the recovery process of fossil animals in some parts of his remains.The development of a focused paleontological method has led to the fact that the correlation analysis began to use in various fields of human activity.
This method is attractive for the processing of statistical data.Correlation analysis in the statistics for the first time used the English biologist and statistician Francis Galton in the late XIX century.In the future development of the method, to measure the closeness of the relationship between the couple and a large number of variables.Correlation analysis has a close relationship with the regression analysis.
occupies a special place in the economy of a correlation analysis.But its use imposes a number of restrictions.First of all, it is a sufficient number of measurements and the data for the study.Practice suggests that the number of observations must exceed 5-6 times the number of factors.The best option is to have the number of cases exceeding the number of factors in several dozen times.In such a case, the law of large numbers, thanks to him, there will be mutually compensating random fluctuations.
should also ensure that the entire set of factor and effective signs obeyed multivariate normal distribution.There are cases where the aggregate volume enough for formal conformance testing normality, then visually determining the distribution law is carried out according to the correlation of the field.If the points are located according to the linear trend, it is quite possible to conclude that the set of initial data satisfy the requirements of a normal distribution.
The initial set of values necessary to monitor the uniformity of quality.
Availability fact more correlation does not give grounds for the claim that an arbitrarily chosen variable is preceded by the appearance of a second or it causes changes in other words, between them there is a strict causal relationship between them, and even a possible effect of a third factor.
applying in practice the results of the analysis based on the correlation of research methods, you can make a number of specific findings on availability, and, most importantly, of the nature of interdependence.It already provides a substantial share of information about the object under study.
