How to Read Statistical Tables in Research Papers
Tables exist because they communicate precise numerical information more efficiently than text. A single table can present dozens of measurements across multiple groups and conditions, with all the associated statistics, in a format that allows quick comparison. The challenge is knowing where to focus your attention, because most tables contain far more information than you need for any single question.
Step 1: Read the Table Title and Footnotes
Always start with the title, which tells you what data the table presents. A title like "Baseline Characteristics of Study Participants by Treatment Group" immediately tells you this is a demographics table comparing groups at the start of the study. A title like "Primary and Secondary Outcomes at 12 Months" tells you this is the main results table.
Then skip to the footnotes at the bottom of the table. Footnotes define abbreviations (SD = standard deviation, CI = confidence interval, OR = odds ratio), explain the statistical tests used, clarify significance markers (typically * for p less than 0.05, ** for p less than 0.01), and note any special conventions. Reading footnotes first prevents confusion when you encounter unfamiliar abbreviations in the body of the table.
Step 2: Understand the Row and Column Structure
Tables are grids, and understanding the grid layout is essential before looking at individual numbers. Column headers typically represent different groups (treatment vs. control, male vs. female) or different statistical measures (mean, SD, p-value). Row labels typically represent different variables being measured (age, blood pressure, test scores) or different time points.
Some tables use hierarchical headers, where a top-level header spans multiple columns. For example, "Intervention Group (n=150)" might span three columns showing mean, SD, and 95% CI. Understanding this nesting is crucial for reading the correct numbers. If you misidentify which column belongs to which group, you will misread the data entirely.
Common table formats include: demographics tables (rows are characteristics, columns are groups), results tables (rows are outcomes, columns are groups and statistics), regression tables (rows are predictor variables, columns are coefficients, standard errors, and p-values), and correlation matrices (both rows and columns are variables, cells contain correlation coefficients).
Step 3: Identify the Key Statistics
Find the row that corresponds to the primary outcome (the main thing the study measured) and the column that shows the comparison you care about. The most important numbers are typically the point estimate (the measured value, mean difference, odds ratio, or correlation), the confidence interval (the range of plausible true values), and the p-value (the probability of seeing the result by chance).
Common statistics you will encounter: Mean (SD) reports the average value with its standard deviation, indicating the spread of individual data points. n (%) reports counts and percentages for categorical variables. OR (95% CI) reports odds ratios with confidence intervals, common in epidemiology. HR (95% CI) reports hazard ratios with confidence intervals, common in survival analysis. Beta (SE) reports regression coefficients with standard errors.
Step 4: Check the Comparison Indicators
Most tables use visual markers to flag statistically significant results. Asterisks are the most common: a single asterisk (*) usually means p less than 0.05, double asterisks (**) mean p less than 0.01, and triple asterisks (***) mean p less than 0.001. Some tables use bold text, superscript letters, or color coding instead. The footnotes will define whatever convention the table uses.
When a table shows multiple comparisons (many rows of outcomes or many pairs of groups), be cautious about significance markers. Without correction for multiple testing, some of those "significant" results are likely false positives. Check whether the footnotes mention any multiple comparison adjustment.
Step 5: Assess Practical Significance
After identifying which results are statistically significant, ask whether they are practically significant. A medication that lowers blood pressure by 1 mmHg with p = 0.001 is statistically significant but clinically meaningless, because 1 mmHg is within the range of normal daily fluctuation. A medication that lowers blood pressure by 15 mmHg is clinically meaningful regardless of the p-value.
Context determines what counts as practically significant. In drug development, a small effect might be important if the drug is safe and inexpensive. In education research, a half-point improvement on a standardized test might be too small to justify the cost of an intervention. In engineering, a 0.1% improvement in efficiency might save millions of dollars at industrial scale. Always interpret the magnitude of results in the context of the specific field and application.
Reading Supplementary Tables
Many papers include supplementary tables that did not fit in the main text. These tables often contain sensitivity analyses (testing whether results hold under different assumptions), subgroup analyses (breaking results down by age, sex, or other characteristics), and additional outcome measures beyond the primary ones. Supplementary tables are not less important than main tables. They frequently contain the most detailed and nuanced information in the paper, and reviewers sometimes request that key analyses be moved to supplementary materials simply to reduce the length of the main manuscript.
When evaluating a study, always check the supplementary materials for tables that test the robustness of the main findings. If the primary result holds across multiple sensitivity analyses and subgroups, you can have more confidence in it. If the result disappears when tested under slightly different conditions, that fragility is important information that the main text may not emphasize.
Common Table Types and What They Tell You
Table 1 (Demographics/Baseline) appears in nearly every clinical study. It compares the characteristics of study groups at the start of the study. If the groups are well-balanced on key variables (age, sex, disease severity), the study's internal validity is stronger. Large imbalances suggest that observed differences in outcomes might be caused by baseline differences rather than the intervention.
Results tables present outcome data, usually comparing groups at one or more time points. The primary outcome should be clearly identified. If the primary outcome is not significant but secondary outcomes are highlighted, the authors may be engaging in selective reporting.
Regression tables show the results of multivariable analysis. Each row is a predictor variable, and the coefficient indicates the strength and direction of its relationship with the outcome after adjusting for all other variables in the model. Larger coefficients indicate stronger relationships, and the sign (positive or negative) indicates the direction.
Survival tables present time-to-event data, often showing the number of participants still at risk at each time interval, the number of events, and cumulative survival or hazard rates. These accompany Kaplan-Meier curves and Cox regression analyses.
Correlation matrices display the relationships between multiple variables simultaneously. Each cell contains a correlation coefficient (typically Pearson r) showing the strength and direction of the linear relationship between the row variable and the column variable. Values near 1 or -1 indicate strong relationships, while values near 0 indicate weak or no linear relationship. These matrices are symmetric, meaning the upper and lower triangles contain the same information, so you only need to read one half.
Read tables systematically: title and footnotes first, then structure, then key numbers, then significance markers. Focus on the primary outcome and always distinguish between statistical significance and practical importance.