如何写好论文中的「研究方法」? 四大覆盖重点方向与误区分析

考量因素

研究架构

(often an introductory paragraph)

• What do we currently know about the research topic?
• What type of study are you conducting (descriptive, analytical, comparative, interventional, observational, etc.)?
• What variables will you use, and which subjects will be exposed to which variables?
• How often and when will the data be collected?
• How can you control all the factors that might affect prediction models and outcomes?
• Do you have to adjust design because of some feasibility issues? If so, what factors?

实验对象 (selection criteria and methods)

• Ethical considerations (all animal or human studies must discuss factors like ethics committee approval of research protocol, informed consent of human subjects, etc.
• Study setting (time, place, etc.), where applicable.
• For living subjects (humans and animals): demographic and clinical conditions, gender, weight, species, age, special characteristics, living conditions, etc.
• Any preparations of subjects made before starting experiments.
• Sampling method, including target population, sampling frame, instruments used, and any stratification, clustering or weighting.
• Recruitment method and its effectiveness, including profiles of any subjects that refused (where applicable).
• For comparative studies: group allocation and randomization procedures.
• Follow-up procedures for longitudinal studies.

资料搜集 (criteria and methods)

• Variables measured (identify only the key predictors and all outcomes of those variables).
• Methods and instruments used to collect data (include information such as an instrument’s manufacturer and model, calibration procedures, and other information necessary to allow others to reproduce your experimental results).
• Bias controls (e.g., blinding procedures).

资料分析

• Descriptive statistics (e.g., means, medians, standard deviations, etc.).
• Inferential methods (include confidence intervals, hypothetical testing methods, and regression models or other modeling procedures used).
• Methods used to address confounding observational studies.
• Interim analysis methods.
• Adjustments made to sampling methods and weighting procedures.
• Missing data.
• Sample size.
• Power determination.
• Software used.