Steps
Step 1
Research Question Definition and Statistical Analysis Plan Design
The core task of this step is to define the research question and design a scientific statistical analysis plan. Transform epidemiological research questions into testable statistical hypotheses, select appropriate statistical analysis methods.
• Define research question: choose specific topics from disease distribution, risk factor analysis, diagnostic test evaluation, prognostic analysis; clarify research purpose (descriptive, comparative, associative, predictive)
• Develop statistical analysis plan: determine data types (continuous, count, ordinal); select statistical methods (descriptive stats, t-test, chi-square, ANOVA, regression, survival analysis); set significance level α=0.05
• Sample size estimation and power: based on study design (cross-sectional, case-control, cohort); use sample size formulas or software (G*Power, PASS); assess power (1-β≥0.8)
Deliverable: Statistical analysis plan (research hypotheses, data types, method selection, sample size calculation, analysis flowchart) | Quality standard: Clear research hypothesis, appropriate statistical method selection, correct sample size calculation, detailed and feasible analysis plan
Step 2
Data Collection and Database Establishment
The core task of this step is to systematically collect research data and establish a standardized database. Design data collection tools, develop data entry specifications, ensure data completeness, accuracy, and consistency.
• Design data collection forms: design CRF (Case Report Form) or questionnaires based on research variables, including basic information, exposure factors, outcome indicators, covariates; define variable coding rules and value ranges
• Establish database: use EpiData, Excel, or SPSS to build database; set variable names, types, labels, value labels, missing value codes; set logical validation and range validation
• Data entry and quality control: use double data entry or post-entry spot check; perform consistency testing (Kappa value, agreement rate); record data cleaning log; save raw and cleaned data
Deliverable: Research dataset (raw database, cleaned database, data dictionary, data cleaning log, quality control report) | Quality standard: Complete and accurate data, standardized database structure, strict quality control, clear variable definitions
Step 3
Descriptive Statistics and Data Exploration
The core task of this step is to perform comprehensive descriptive statistics and exploratory data analysis, understand data distribution characteristics, discover patterns and anomalies. Descriptive statistics is the first step of statistical analysis.
• Continuous data description: calculate mean, median, standard deviation, interquartile range, max, min; determine distribution type (normal, skewed); draw histograms, boxplots, Q-Q plots
• Categorical data description: calculate frequency, proportion, composition ratio, relative ratio; draw bar charts, pie charts, Pareto charts; describe distribution characteristics of categorical variables
• Data exploration and outlier testing: perform normality tests (Shapiro-Wilk, Kolmogorov-Smirnov); identify outliers (3σ rule, boxplot method); analyze missing value patterns; explore preliminary relationships among variables (scatterplot matrix, correlation matrix)
Deliverable: Descriptive statistics report (statistical description tables, distribution plots, exploratory analysis charts, data quality assessment) | Quality standard: Comprehensive and accurate description, standard and clear charts, clear distribution characteristics, reasonable outlier handling
Step 4
Inferential Statistics and Association Analysis
The core task of this step is to test research hypotheses using inferential statistical methods and analyze associations among variables. Select appropriate statistical test methods based on study design and data types.
• Group comparison: select test methods based on data type and design—continuous data use t-test (independent/paired) or ANOVA (one-way/two-way/repeated measures); categorical data use chi-square, Fisher exact test, or rank sum test
• Association analysis: calculate correlation coefficients (Pearson product-moment, Spearman rank); perform simple and multiple linear regression; calculate association indicators (OR, RR, HR) with 95% confidence intervals
• Confounding control: use stratified analysis (Mantel-Haenszel method), multivariable regression (logistic regression, Cox proportional hazards model); test for interaction; draw forest plots to present results
Deliverable: Inferential statistics analysis report (hypothesis test results, association strength indicators, regression analysis tables, forest plots, hypothesis test conclusions) | Quality standard: Correct application of statistical methods, reasonable result interpretation, proper confounding control, evidence-based conclusions
Step 5
Statistical Results Reporting and Paper Writing
The core task of this step is to report statistical analysis results in a standardized manner and write a complete research paper or statistical analysis report. Follow statistical reporting guidelines (STROBE, CONSORT).
• Result reporting standards: follow reporting guidelines corresponding to study type (STROBE for observational, CONSORT for RCT, STARD for diagnostic tests); report effect sizes with 95% CI; provide exact P-values
• Create statistical charts: produce statistical tables (three-line tables) and graphs (scatter plots, bar charts, survival curves, ROC curves) following journal standards; clearly label axes, legends, error bars, statistical symbols
• Write analysis report: compose complete statistical analysis report or academic paper including abstract, introduction, materials and methods, results, discussion, conclusion, references; discuss difference between statistical and practical significance; analyze study limitations
Deliverable: Complete statistical analysis report/research paper (all statistical results, standard charts, discussion), statistical analysis code and data files | Quality standard: Complete report structure, detailed statistical method description, standard result presentation, in-depth and thorough discussion