For a long time, scientific research has moved from simple qualitative analysis to detailed quantitative analysis, and researchers have to face a large number of data analysis problems. The statistical analysis results of scientific research data directly affect the results analysis of papers. In medical research writing, the method of experiment design directly determines which statistical method to use for data, because each statistical method requires data to meet certain preconditions and assumptions, so the paper should consider which statistical method is more reliable in the future when designing experiments. There are many mistakes in medical statistical methods, the most important of which is that the statistical methods are inconsistent with the experimental design, which leads to unreliable statistical results. Below, the editors of medical journals listed some common avoidable problems and mistakes:

Open Baidu APP to see more high-definition pictures.

First, the data statistical analysis method is used incorrectly or improperly. In medical papers, the most common mistake is that when the experimental design is multi-group research, it is necessary to analyze the variance of the data, and the authors always use the mean value test of two samples.

Second, the statistical method is not clear. For the same medical paper, different data should be processed by different statistical methods, which requires the author to clearly describe what statistical method is used for each statistical value. However, in many medical papers that use a variety of statistical analysis methods, the author often simply lists the statistical methods used in the paper as a whole, and does not explain the specific statistical methods for each data result analysis, which makes it difficult for readers to confirm what data analysis methods the author used for a specific result.

Three, statistical tables and charts are missing or duplicate. Statistical tables or charts can intuitively let readers know the statistical results. A good statistical table or chart should be independent, that is, the author can infer the correct experimental results from the statistical table or chart even without looking at the content of the article. However, some medical papers simply pile up a lot of statistical data, lacking intuitive statistical charts or tables; Or although statistical tables or charts are listed, there are many omissions in the tables or charts, and it is difficult for readers to extract too much useful information from them.

In addition, in order to increase the length of articles, some authors also list statistical tables and charts, resulting in unnecessary waste and repetition. The advantage of statistical tables is that they are detailed and easy to analyze and study various problems. The advantage of statistical charts (especially bar charts) is that they can directly reflect the quantitative differences of variables.

The two most common mistakes in interpreting statistical results in medical papers are over-reliance on P value (a decreasing indicator of the reliability of results) and avoidance of negative results. The reason for the former error is that some authors misunderstand the meaning of P value and confuse the statistical significance of data with the clinical significance of research. Therefore, medical researchers must be careful not to rely solely on statistical values to draw arbitrary conclusions, and must combine statistical results with clinical practice, so as to avoid similar mistakes.

As for avoiding negative results and only providing positive results, it is because many authors can't get rid of a one-way mindset when studying design, that is, subjectively identify some expected results and conclusions first. When summing up the reasons for a certain result, we should draw a perfect conclusion from the experiment in one direction, especially if this conclusion may be very meaningful to the actual situation. This mindset places too much emphasis on the significance of statistical differences, sometimes deliberately avoiding the unobvious results of reporting differences, not thinking about and exploring the reasons and significance of unobvious differences, but ignoring some major scientific discoveries.