Validation is carried out at all stages of collection and processing of statistical data: filling in forms of statistical monitoring, reporting, development of statistics and preparation of analytical tables.
Chance of data is checked by means of an arithmetic or logic control. Arithmetic control is based on the use of the relationships between various indicators reports and checking all general indicators and harmonization of those which are derived from each other.
The value of correlation coefficient cannot be the measure of validity of correlation between features. This parameter depends on the number of degrees of freedom. For larger n the reliability of communication at the same value of the correlation coefficient is higher.
The value of correlation and regression are considered reliable if they exceed their mistakes in a certain amount of time, depending on the sample size. Criteria compared with the standards of reliability values for the Student table set number of degrees of freedom and probability threshold infallible predictions.
,, where
- r – correlation coefficient;
- n – the number of observations (the number of pairs);
- t – Student’s criterion (or validity).
Thus, to defend myself and my program I would like to perform the validity analysis for the given correlation coefficient. If the validity of the coefficient will be low, we have enough evidence to claim that the correlation coefficient is not significant and we have to check our data set and find a possible issue which has leaded to such result. The most likely reason that the data set is not representative (for example, students were not chosen randomly).
