For psychological empirical research, data analysis is almost necessary, especially various statistical significance tests. If the parameter test is used, the data needs to meet some conditions. For example, for the studied variables, the population from which the sample comes must conform to the normal distribution, and the sample data must also conform to the normal distribution. As we all know, many human attributes, such as height and intelligence, are in line with the normal distribution. Therefore, in terms of the normal distribution requirements of parameter test, it is often for samples.

Whether the sample data conform to the normal distribution can be determined by test. If the test results show that the sample data conforms to the normal distribution, the parameter test to be done can be carried out. If the test results show that the sample data does not conform to the normal distribution, then the nonparametric test can be used, or the data can be converted to make the data conform to the normal distribution for parametric test. Because the effectiveness of nonparametric test is low, data conversion is a more preferred choice in this case.

Many researchers will refer to the existing research methods and convert the data when necessary. However, they may not understand why the data can be converted, and even feel confused: what is the matter that the data that clearly does not conform to the normal distribution should be forcibly converted to the data that conforms to the normal distribution? Is it openly allowing tampering with data? Therefore, some researchers know data conversion and have also carried out data conversion. However, it is speculated that they still do not know the principle of data conversion. In particular, if one transformation does not turn the data into a normal distribution, another transformation can be carried out. This is even more disturbing. How is this different from manipulating data?

Of course, data transformation is neither tampering with data nor manipulating data, but transforming the measurement scale of data. Generally speaking, data transformation is to transform the measurement unit of research attributes. Pupils have learned relevant knowledge and done relevant exercises in mathematics class. For example, 1 meter = 3 feet, 4 tons = 4000 kg, 5 hours = 300 minutes. When the ruler is used as the unit, it is assumed that there are three measurement data, which are 3.45, 8.28 and 11.94 respectively. If the meter is used as the unit, the corresponding three measurement data are 1.15, 2.76 and 3.98 respectively.

It is not difficult to see that when the ruler is used as the unit, the number is large, and the gap between the numbers is also large; When the meter is used as the unit, the number is small and the gap between the numbers is small. More importantly, although the numbers of the two sets of data are different, their corresponding physical length is the same. In other words, data is important, but specific numbers are not so important. Therefore, for the length of an object, it is possible to express it with 3.45, 1.15, or even 11.50.

Meanwhile, for the above set of data, if the number 3.45 becomes 1.15, then 8.28 and 11.94 must also become 2.76 and 3.98. Obviously, this digital conversion is systematic, or systematic conversion according to certain rules. Although the data has been converted into different numbers, the number of physical attributes corresponding to the data is still the same as before. From the perspective of operation, it was originally measured with a ruler in the unit of ruler, but now measured with a ruler in the unit of meter.

Quantitative research in empirical research is that researchers measure specific attributes, obtain data, statistically analyze the data, get results, and then draw conclusions. Although every link in the research process is important, some links are more critical. For example, measurement is one of such key links. After all, the data of statistical analysis comes from measurement, and the research results and conclusions are based on statistical analysis.

When it comes to measurement, it is easy to think of holding a ruler to measure the length of an object. Measuring the length with a ruler may be the first case of measurement. Obviously, people need to measure not only length, but also weight; Not only physical attributes but also psychological attributes need to be measured. For the properties of the study, the results will be obtained after measurement, that is, numbers. Thus, the so-called measurement is the process of giving numbers to the attributes of things according to certain rules.

It can be seen that the rules on which the measurement is based are basic. It is advisable to call the measurement rules scale. When measuring, for the same attribute, different scales are used, and the numbers obtained are naturally different.

Some scales have been known or familiar with their specific meanings, such as common scales of length, weight and speed. In fact, there are more scales that people are not familiar with or do not know at all. In particular, some scales are more complex, which may be the ratio of two or more attributes, such as mach. Researchers will also formulate or define new scales in specific research.

The point here is that researchers may or may not use the appropriate scale when measuring. Even, for the measured attributes, the commonly used scales are not appropriate, and the appropriate scales have not been found or defined. Therefore, the corresponding numbers are not particularly effective when using the data measured with the existing scales. For example, they do not conform to the normal distribution and cannot be tested for parameters. At this time, the natural idea is to change the measurement scale and make the measurement numbers conform to the normal distribution, so as to carry out more effective parameter test.

Statisticians have systematically studied the law of data conversion and put forward some common conversion methods, such as standard fraction conversion, square root conversion, logarithmic conversion, sine conversion, cosine conversion, etc. It can be seen that data conversion usually changes the measurement scale from large to small, correspondingly, changes the gap between data from large to small, and changes the data that does not conform to the normal distribution into the data that conforms to the normal distribution.

In short, understanding the basic meaning of measurement, especially the problem of measurement scale, will truly understand the principle of data conversion, so that researchers can safely and flexibly carry out data conversion according to needs.