The selection of statistical data significance test method is a common problem in medical papers, and the reason for rejection is often improper selection of significance test method. Such as t test, u test, χ2 test, etc. Although each has its own scope of application and requirements, there are also similarities. The author can choose one or several test methods according to the type of statistical data. However, when the author obtains one, two or more sets of data, it is very important to choose which significance test. Different data types have different statistical indicators and statistical test methods, as shown in table 1.
In medical biology research, many indexes obey normal distribution (U distribution), but with the increase of sample content or degree of freedom, T distribution, χ2 distribution and F distribution tend to normal distribution, as shown in Figure 1 and Figure 2.
The statistical methods involved in the articles in the Chinese Journal of Trauma, Vol. 12,No. 1 ~ 6 and its supplement (Table 2) show that the normal distribution is widespread and universal.
So when the author obtains the data, he should conduct a normal test first? Norm is standard normal distribution (or nearly normal distribution) or does not belong to normal distribution. The author first recommends the probability unit method.
When the statistical data belongs to normal distribution or approximate normal distribution, what is the difference significance test method? According to its application conditions, it can generally be selected according to Table 3.
Main points for attention in the application of significance test: (1) Before significance test, we should pay attention to the representativeness and comparability of samples. (2) When the test results are close to the significance limit, it is necessary to consider whether there is really no difference in many aspects; Or the number of observed cases is not enough, and it is necessary to increase the sample cases. Improper use of test formula can be confirmed by other tests. (3) The χ2 test of the proportion of multiple groups of samples shows that there are significant differences, which can only show that the proportion of multiple groups is different or not exactly the same, but can't determine which proportion is different. Further significance test is needed to know whether the ratio of the two samples is the same.
Table 1 Relationship between statistical indicators of different data and general inspection methods.
Statistical test method for statistical indicators of data types
Measurement data mean, standard deviation t test, f test, etc.
Counting data rate, composition ratio χ2 test, etc.
Semi-quantitative data rate, composition ratio rank sum test, Ridit analysis
Table 2 Chinese Journal of Trauma Vol. 12,No. 1 ~ 6,
Frequency of significance test method in supplement
Test method application times test method application times
T test 27 linear correlation and regression analysis 5
χ2 test 16 fitted linear regression 1.
F test 24 correlation analysis 6
Q test 2 nonparametric statistics 4
U test 1 does not mean method 6.
Table 3 Selection of common significance test methods
Comparative significance test of statistical data
T-test for comparison between small sample mean and population mean
Small sample mean comparison t test, f test
Average value of two or more large samples
Comparison between u test and t test of population mean
Comparison between u-test and t-test of large sample mean
Paired t-test of paired measurement data
Comparison of two ratios U test and χ2 test
Multi-sampling rate comparison χ2 test
Pairing of two attributes of counting data
Correlation analysis and its difference comparison χ2 test