Error analysis of the application of medical statistical methods in statistical papers 1.
I. Introduction
Due to the complexity and systematicness of medical research, it is often necessary to apply rigorous statistical methods. Because some authors lack a deep understanding of the application of statistical theories and methods in medical research, the phenomenon of wrong application of statistical methods in medical papers occurs from time to time. Errors in the application of statistical methods directly lead to errors in statistical results. Such as statistical charts, statistical indicators, statistical significance tests, etc. Therefore, the correct application of statistical methods and the correct description of the results will help to improve the quality of a single paper. The errors in the application of statistical methods in medical papers and their common results are analyzed as follows.
Second, the application of statistical methods in medical papers
Abstracts of medical papers are highly concentrated [1], which mainly consists of objectives, methods, results and conclusions. General requirements indicate the main statistical methods, statistical research results and P value. The quality of medical papers can often be judged by the statistical results of abstracts. The choice of statistical methods and the expression of results directly affect the scientific research level of personal works.
(1) materials and methods
In the text, the selection, application and statistical significance of statistical methods must be clearly explained in the materials and methods section. Through the description of statistical methods, readers should be clear about the statistical design idea of this book. The material part should clearly explain the source of the sample or case, the inclusion and exclusion criteria, the sample size, the setting conditions of the research group and the control group, retrospective or prospective research, investigation or experimental research and other general information related to the research. The purpose is to explain the rationality and reliability of the application of statistical methods, so that others can do relevant research with repeatability. The method part should describe in detail the different treatment processes, specific observation indexes and measurement techniques adopted by the research group and the control group, which is comparable and scientific.
The method part also introduces the statistical analysis method and the statistical software used. Different data processing should adopt different methods, and the counting or measuring data, the comparison between two or more groups, and the correlation research of different processing factors should be clearly explained. Commonly used are T test of measurement data between two groups, F test of measurement data between multiple groups, chi-square test of counting data, correlation analysis and regression analysis among different factors. Some genetic research methods also have special statistical methods. Here we should briefly explain and give reference, and also briefly describe the principle of statistical methods. Statistical software should clearly indicate the software name and version number, such as FBAT 1.7.3 version based on family data research.
(2) The result of the paper.
The results of the paper should show the statistics obtained by statistical methods [2], and when many statistical indicators are used, they are often described separately. When the number of groups is large, statistical charts are needed to accurately express the statistical results. The accuracy of the data is not only related to the accuracy of the measuring instrument, but also to the average value of the sample itself. The unit of the obtained value generally adopts the principle that the nearest average value is divided by three. The significant figures of mean and standard deviation should be consistent with the original data. The standard deviation or standard deviation sometimes needs to be increased by one digit, and the percentage generally keeps one decimal place. In statistical software, the accuracy of the analysis results is often high, so rounding is generally used to make it close to the actual situation of the experiment, otherwise the credibility and readability of the paper will be reduced.
Results The statistical tables in the parts are unified? Third line? Data types, such as mean and standard deviation, should be indicated in tables and table topics. The values in the table should be placed in the order of rows and columns, neat and beautiful, and there should be no wrong rows. It is necessary to clearly mark the number of observed cases and the test statistics obtained. Statistical charts can directly express the research results, such as the scatter plot of regression and correlation analysis can show the dispersion degree of a single value. The figure shows the results of repeated measurement of the average values of individuals in different groups over time or under different conditions. The error histogram is drawn by the average addition and subtraction standard error, which describes the confidence interval of 67%, not 95%. It is suggested to use the 95% confidence interval in the error bar chart.
As for statistics, the average and standard deviation are generally used, and the average cannot be used alone. When using the average, it is necessary to clarify the standard deviation of the variation index or the standard error of the accuracy index. Regarding the percentage, the determination of denominator must be logical. If the sample is too small, the denominator will be too small and the percentage will be too large. In the comparison of percentages, to write down the different changes between them, you can use chi-square test.
1. In the results of hypothesis testing, it is common to write only P value, which sometimes misleads readers and hides calculation errors. Therefore, writing specific statistical values, such as F value and T value, can enhance the credibility. For descriptive statistics such as ratio, correlation coefficient and mean value, it is necessary to clearly explain that statistical tests have been carried out and list the results. Generally, the p values are 0.05 and 0.0 1, and the calculation results need specific p values, such as P=0.23 or P=0.02.
2. When discussing papers, the results produced as statistical methods are often used as the author's main viewpoints to support his scientific assumptions, which is very important to correctly explain the statistical results. A large value of p indicates that there is no difference between the two groups, which is a high probability event, and a small value of p indicates that the probability of no difference between the two groups is very small. When p
False correlation between variables is related to the change of variables with time, and statistical correlation does not mean that there must be causal relationship between variables. The determination of causality should be based on different professional knowledge and research methods. Using regression equation to analyze, when there is a significant relationship between two variables, it is still not very accurate to infer the dependent variable from the independent variable. Correlation or regression coefficient can not predict the accuracy of the results, but can only predict a credible interval. Diagnostic tests are applied to diseases with low incidence in the population, and the diagnostic sensitivity and specificity of diseases are not very certain. ? False positive rate? With what? False negative rate? According to actual needs, different requirements are not consistent. When the disease prevalence rate is low, it is normal to have false positives. The diagnosis of the disease must be combined with clinical symptoms and signs. Therefore, the calculation methods of these two rates must be clear.
Third, analysis of common errors in the application of statistical methods in medical papers
( 1)? Materials and methods? Common mistakes in the application of statistical methods
? Materials and methods? The common problems of statistical methods are: the selection of samples or the description of the source and grouping of research objects are few or too simple. For example, simple random grouping of clinical cases did not describe the method of random grouping, whether it was double-blind and double-simulated, and no blank control group was set up. After grouping, the descriptions of gender, age and education level were not statistically tested, and the special statistical methods were not explained in detail; The description of the randomization principle of animal experiment grouping is too simple, and there is no provision for complete randomization, pairing or stratified randomization. Statistical analysis methods do not explain the analysis software used, and some only explain the analysis software used and do not explain the statistical methods used in the software; The sample size is too small to explain the reason.
(2)? And the result? Common mistakes in the application of statistical methods
1. The expression of the result obtained by applying the correct statistical method may be incorrect. For example, the accuracy requirements of the data mentioned above. The common mistakes in medical papers are: the statistical indicators such as mean, standard deviation and standard error are different from the decimal places that the original data should retain; There are too many decimal places in statistical indicators such as ratio, number of cases, ratio, ratio and relative risk; The incidence rate, prevalence rate and prevalence rate of rare diseases are not well selected, resulting in no integer results; Correlation coefficient, regression coefficient and other indicators have too many or too few decimal places; Some commonly used inspection statistics such as F value and T value do not meet the requirements.
2. When analyzing and calculating statistical indicators, counting data and measurement data are generally used to distinguish them. Three-line table is often used to measure data, which is described by mean and standard deviation on the premise of approximately obeying normal distribution. If it does not conform to the normal distribution, it can be approximated by logarithm or other processing methods, otherwise it can only be described by indicators such as median and interquartile spacing. It is common in medical papers that the calculated data have not been tested by normal distribution, which affects the authenticity and credibility of statistical results. For commonly used counting data indicators, such as ratio and composition ratio, there is a general problem of small sample size. Using ratio to describe will affect the reliability of statistical results, and using absolute number to explain will be more objective. There are also some documents that misuse the composition ratio as ratio, which is also not desirable.
3. When judging an index of clinical curative effect, the difference of average curative effect between the two groups does not mean that each individual in the two groups is effective or invalid, but must be calculated by calculating the effective rate. For example, comparing the curative effect of a drug on diabetes, after taking the drug for one week, the blood sugar of the study group and the control group decreased by 6.7 respectively? 2.4 and 1.2? 0.6 (P = 0.000 1). According to the curative effect of fasting blood glucose below 7.7mmol/L, the effective rates of the study group and the control group were 75.6% and 65,438 02.4% respectively. Although the average curative effect varies greatly, it should be noted that this drug is ineffective for some patients (24.4%). For the statistical analysis results of hypothesis test results, the expression of p value advocates reporting accurate p value, such as P = 0.0 15 or P = 0.32 1. At present, all statistical analysis software can automatically calculate the accurate P value. For example, commonly used SAS, SPSS and so on. As long as the original data is provided, the t value, the f value and the corresponding degrees of freedom can be calculated, and the accurate p value can be obtained.
Fourth, knot
Improving the use quality of statistical methods in medical papers is a long-term and arduous task that the editorial department should pay attention to [3]. Whether the use of statistical methods and the expression of statistical results in medical papers are correct not only reflects the scientific and rigorous nature of the papers, but also plays an important role in improving the overall academic quality of journals and promoting the development and dissemination of medical science [4].
References:
[1] Li Jingwen, Lu,. Influence of adding abstracts of key articles in medical journals on the cited frequency [J]. acta editologica, 20 1 1(23).
Chen Changsheng. Expression and interpretation of statistical results in biomedical papers [J]. Journal of Cell and Molecular Immunology, 2008(24).
[3] Pan Zhiming. Quality and ability of editors of compound medical sci-tech periodicals in the new period [J]. Research on China Sci-tech Journals, 20 1 1 (22).
Analysis of graduation status of statistics major and its countermeasures.
Undergraduate graduation thesis is an important part of the talent training plan in colleges and universities, the last important teaching practice link in the undergraduate teaching process, and an important basis for awarding bachelor's degrees. [1, 2] However, compared with other teaching links, graduation thesis has not received enough attention, which leads to some problems in this link. [3] This paper will take the graduation thesis of statistics major of Central University for Nationalities as an example, and on the basis of analyzing its present situation, find out the problems and put forward corresponding suggestions.
The statistics undergraduate major of Minzu University of China was established in 2003, and now there are six graduates. Through the efforts of colleges and universities, statistics, as a new major, has made rapid development, and the students trained have good professional ability and comprehensive quality. Nearly 40% of the students continue their graduate studies, and most of the employed students are in jobs corresponding to their majors, and the employment rate has been around 85%.
The undergraduate graduation thesis link is 6 credits in the training program. In the seventh semester, students began to choose tutors and determine the topic of graduation thesis. After six semesters of systematic theoretical study, students majoring in statistics have basically mastered the basic theories and methods of statistics, and have correct statistical ideas and strong application ability of statistical software, as well as comprehensive ability to solve practical problems, document retrieval and data query by using the theories and methods they have learned. The writing of undergraduate graduation thesis is an important process for students majoring in statistics to further deepen and sublimate the above-mentioned foundation and ability, thus cultivating students' innovative ability and practical ability, and further enriching and improving students' knowledge, skills and quality. It is also an important symbol to measure the quality of teaching and the level of running a school. Therefore, it is necessary and meaningful to study how to improve the quality of graduation thesis. [4]
Firstly, the quality of graduation thesis of statistics major is analyzed.
Judging from the completion of the thesis, each graduation thesis can basically meet the requirements of thesis teaching. By analyzing the graduation theses of the graduates majoring in statistics in 2007-20 1 1, it is found that the passing rate of graduation theses is 99. 18%.
Judging from the types of graduation thesis research, it can be divided into two categories: theoretical research papers and empirical papers. The characteristics of theoretical research papers are to summarize and discuss the existing statistical theoretical problems, express the results of theoretical research, or apply theory to analyze and explain practical problems and put forward their own ideas; The empirical paper is mainly manifested in the statistical investigation, statistical information management and quantitative analysis of economy, management, finance, medicine, biology, engineering, environment and other fields by using statistical theories and methods for specific practical purposes or goals.
From the analysis of the knowledge scope of the thesis, most of the students' papers are problems of statistics major, and very few are problems of other branches of mathematics. From the analysis of the graduation thesis of previous students majoring in statistics in Minzu University of China, it is found that the problems in learning other branches of mathematics account for 6.50% of the total, mainly including: First, the applied research of other disciplines (mathematical analysis, ordinary differential equations, operational research, spatial analytic geometry, etc.). ), accounting for 1.63% of the total. Second, the education of mathematics majors and the research of mathematical thinking account for 4.87% of the total. Statistics graduation thesis accounts for the vast majority, accounting for 93.50%. The topics are extensive, mostly social hot issues, involving economic, social, medical and health, educational development, tourism, infrastructure construction and other fields. Due to the influence of the school's humanistic environment, a large proportion of students have made statistical analysis on the economic, social and people's livelihood issues in minority areas, accounting for about 30.08% of the total number of students. The analysis methods used mainly focus on sampling survey, regression analysis, multivariate statistical methods, cluster analysis, discriminant analysis and other commonly used statistical methods.
In addition, statistical analysis shows that students' scores are generally high, and graduation thesis of statistics major, especially empirical thesis, has the characteristics of extensive use of background introduction and statistical software analysis results. So some papers are not innovative and academic, but they are very long. Compared with the graduation thesis scores of other majors in the College of Science, its average score is relatively high, about 80.42 points.
Second, the problems in the graduation thesis of statistics major
The quality of graduation thesis is related to the training specifications and objectives of undergraduate talents, which directly reflects students' learning achievements at the undergraduate stage and is an important basis for measuring teaching level, students' graduation and degree qualification certification. Through the concrete analysis of papers and evaluation results, it is found that there are many problems in students' thesis writing in terms of innovation, theoretical depth and common sense. The specific performance is as follows:
1. Not innovative enough.
Students' graduation thesis is characterized by little theoretical research, mostly empirical papers, and mostly simple statistical methods, lacking innovative research and thinking. Judging from the graduation theses of students majoring in statistics in previous years in Minzu University of China, only 4.89% are theoretical research papers, and the proportion with empirical papers is 1. 19.45, which shows the characteristics of students avoiding the important and neglecting the innovation in the process of selecting topics for graduation theses. For example, a certain number of students choose every year? China GDP per capita forecast? The main content of this paper is to use one or more methods such as ARMA, grey prediction or trend extrapolation to model and analyze the time series data simply. The paper does not comprehensively analyze the meaning of the index itself and the social and economic situation at home and abroad. This method is simple and applicable, and the proportion of papers is large.
2. The topic is too big, the content is vague, there is a lack of in-depth research, and there are plagiarism and patchwork.
When choosing research topics, some students often can't locate accurately according to their own professional knowledge structure and social practice, but blindly choose some social hot issues that are too big and empty, so it is difficult to see the essence of the problems to be studied. If some students do CPI research, they don't have a deep understanding of the nature of the problem, but only collect some documents, so it is difficult to put forward their own views or research angles, which leads to excessive accumulation and dispersion of materials, and the paper becomes a simple patchwork of some materials. Some papers focus on a social and economic problem. The theme of the paper is only to analyze the existing data by using simple statistical methods, without specifically discussing the quality and reliability of the data and the applicability of the method, and without analyzing the conclusion in combination with the social and economic reality, which leads to the low quality of the paper.
3. Relatively advanced analytical methods are seldom used.
The cutting-edge analytical methods are rarely used, and the statistical analysis ability has not been substantially improved through the writing of graduation thesis. The statistical methods used in students' papers mainly focus on regression analysis, cluster analysis, discriminant analysis and correlation analysis. Among them, the proportion of regression analysis is very large, about 4 1.46%, and the proportions used by other statistical methods are: cluster analysis 5.69%, discriminant analysis 7.32%, correlation analysis 14.63%, and multivariate statistical method 2.44.
4. There are some problems in the writing of the thesis, such as unreasonable structure, no introduction of relevant research, unclear expression of innovation, incorrect labeling of references and so on.
Judging from the students' graduation thesis, the thesis writing is not standardized and the professionalism is poor. There are some problems such as nonstandard form, unreasonable structure, vague theme, disordered papers, serious colloquialism and poor readability.
Third, the analysis of the causes of problems
In view of the above problems, the Statistics Department found that the main reasons for the above problems in graduation thesis include the following aspects through detailed examination of graduation thesis and discussion between teachers and students:
1. Students don't pay enough attention to papers.
Some students are too busy taking the postgraduate entrance examination to study their graduation thesis, and some students are too busy going out to find a job internship to write their thesis seriously. It is difficult to ensure the necessary time for writing a paper, and students can cope with it, so it is impossible to ensure the depth of the paper. In addition, some students think that graduation thesis is only a teaching link and has nothing to do with the quality of postgraduate entrance examination. There is a fluky psychology that teachers will let themselves go as long as they write papers, and they have not paid enough attention to it ideologically.
2. Lack of targeted guidance from tutors.
The number of graduates brought by tutors is too large, which leads to the overload of tutors' workload and can't guarantee the quality of graduation thesis for each student. As a result, some students' papers are not standardized and there are many problems in the papers.
3. Students' professional training is not enough
Most undergraduates have not experienced the training of thesis writing, and their writing level is low. They don't know the standard and format of academic papers, and they don't know how to conceive articles, organize materials and arrange structures from the perspective of scientific research, which makes the views expressed by a considerable number of students' graduation papers inaccurate and unclear, and the arguments can't support them well. In addition, in order to complete the task, some students directly pieced together the information searched in the network without thinking, which made the content unsystematic and chaotic.
Fourthly, suggestions and practices to improve the quality of graduation thesis.
1. Strengthen the publicity of the importance of graduation thesis and improve students' attention.
Strengthening the understanding of the importance of graduation thesis is helpful to improve the quality of undergraduate graduation thesis. Through lectures, classroom teaching and other forms, students can realize that the practicality and comprehensiveness of graduation thesis are irreplaceable by any teaching link. It is an effective way to improve the ability of finding, analyzing and solving problems, and it is also an indispensable link to improve one's comprehensive quality. [4] Make students realize the importance of graduation thesis ideologically and put more energy into graduation thesis design.
2. Combine the topic selection with teachers' scientific research projects to improve the innovation of the paper.
When choosing a topic, in order to give full play to students' subjective initiative, students can make a project plan according to their own characteristics, negotiate with their tutors and combine their research direction. Teachers majoring in statistics generally apply for projects with high theoretical and innovative requirements, such as the National Natural Science Foundation and the National Social Science Foundation, and many teachers also host or participate in related applied research projects. Most application projects need field research (as well as questionnaires and data analysis) or a lot of data analysis and modeling. Guiding students to participate in such projects and design their own undergraduate thesis can stimulate their enthusiasm for scientific research and innovation potential. In addition, encouraging and guiding some students with good grades, such as letting students with graduate degrees participate in teachers' research seminars or research groups, and selecting some theoretical problems with certain difficulties for research, can let students know the development direction and latest trends of this discipline. In recent two years, more and more students, especially those who are qualified for postgraduate entrance examination, can devote themselves to the writing of projects and graduation thesis in the last semester of their senior year.
3. Pay attention to the usual practical teaching links, and cultivate students' practical ability, ability to find and solve problems.
In order to improve students' interest in learning and their ability to analyze and solve problems, a variety of social practice activities have been widely carried out, so that students can get in touch with practical work related to their majors as soon as possible and realize how to combine theory with practice and understand the actual business of their majors, so as to improve their autonomous learning ability and strengthen their mastery of professional knowledge. According to the actual situation of the school, students are actively encouraged to participate in the national mathematical modeling competition for college students at the school level and the national level in their sophomore and junior years. Central University for Nationalities undergraduate research and training program project? 、? Beijing University Students' Scientific Research Project? And then what? National college students' innovative experimental plan? . The application and implementation of projects and the writing of research reports are all good exercises for students. At present, the participation rate of undergraduate students majoring in statistics is above 70%. In addition, the establishment of professional practice base can improve students' ability to analyze and solve practical problems by using professional knowledge. The design and implementation of these links effectively ensure the level and quality of students' undergraduate thesis.
4. Strengthen students' training in writing scientific papers.
At ordinary times, we should strengthen the standardization of classroom homework and cultivate students' writing ability of scientific papers imperceptibly. Through the usual practical activities, such as students' mathematical modeling, college students' innovative practice, etc., students' paper writing ability is improved.
5. Implement incentive measures to stimulate students' interest and initiative.
For students who participate in practical projects, the college encourages tutors to reward them in the form of labor fees according to the completion, and actively encourages students with excellent graduation thesis quality to submit and publish. In addition, it is necessary to standardize the defense procedures and grading standards, establish an excellent graduation thesis instructor and an excellent graduation thesis reward system, form a positive guiding role, and fully mobilize the enthusiasm of instructors and students.
6. Strengthen teachers' sense of responsibility and establish a perfect mechanism.
Strengthen the process management of students' graduation thesis, strictly implement the exam from the time when the topic is due, and guide teachers to strictly check. In order to ensure the communication between teachers and students, schools can improve the separation of teachers and students by establishing a feedback mechanism between teachers and students, providing convenient communication channels for teachers and students, and setting up an appropriate teacher incentive system. At present, Minzu University of China has an extra class allowance for teachers to guide undergraduate graduation thesis.