SOST10021 Chap.9 Tabulation and Data Analysis
Tabulation and Data Analysis
Tabulation and Data Analysis develops a reproducible route from categorical counts to group comparisons. A cross-tabulation can show counts, row percentages, column percentages or total percentages, all of which answer different conditional questions. The research sentence should decide the denominator before software is used.
Percentage points express absolute gaps, while ratios express relative comparison and depend on reference direction. Missing categories, sparse cells, weights and recodes must remain visible. The chapter treats software output as an auditable sequence of variable checks, frequencies, table commands, totals and plain-language interpretation.
What this chapter covers
- 01
Cross-tabulations
- 02
Row percentages
- 03
Column percentages
- 04
Percentage points
- 05
Relative risk
- 06
Missing denominators
- 07
Reproducible commands
Worked example · free
Move from cell counts to a bounded comparison
- 2Group A is 40 divided by 500, or 8%.
- 2Group B is 50 divided by 1,000, or 5%.
- 2The absolute difference is 3 percentage points and relative risk is 1.60.
- 2Describe the association without causal overreach and preserve both denominators. This weighting is original.
Key terms
- Row percentage
- A row percentage divides each cell by its row total to answer a within-row conditional question.
- Percentage point
- A percentage point is the arithmetic difference between two percentages.
- Relative risk
- Relative risk divides one group proportion by a stated reference-group proportion.
Tabulation and Data Analysis FAQ
How do I choose row or column percentages?
Complete the sentence 'among each group' and identify the grouping variable. Percentages should sum within the groups whose outcome prevalence you want to compare. Table orientation is secondary; the research question determines the conditional denominator.
Why report counts with percentages?
Counts reveal the data volume behind each percentage, allow reverse-checking and expose sparse cells. Percentages support comparison across unequal group sizes. Showing both reduces the chance that a large count is mistaken for high prevalence or a small base is hidden.
What makes an analysis reproducible?
Record the dataset version, inclusion rules, missing values, recodes, weights, commands and rounding. Another analyst should be able to rebuild the table and identify where a changed result entered. A saved workflow is part of evidence quality.
When should categories be combined?
Combine categories only when the substantive meaning remains defensible and sparse cells or disclosure risks justify the loss of detail. Preserve the original coding, record the recode, and compare results before and after combining. Convenience alone is not enough when grouping conceals an important inequality.
Exam move
Turn each table question into a conditional sentence before opening software. Decide whether the comparison is outcome prevalence within groups or group composition within an outcome, then request the corresponding percentages. Recalculate several cells from counts and margins and verify that percentages sum in the intended direction.
Practise converting the same values into a percentage-point gap, percent change and relative risk, labelling the reference group and explaining why the scales differ. Show missing, refused and not-applicable cases separately before choosing an exclusion rule. Inspect sparse cells for instability and privacy concerns, and preserve original categories in the audit trail when recoding.
If weights are used, distinguish weighted population estimates from unweighted respondent counts and use design-aware uncertainty. Save commands or scripts with the data version and filters. Build plain-language conclusions that include numerator, denominator, direction and period, then add the design boundary. Avoid causal wording for a descriptive cross-tab.
Keep error categories for wrong percentage direction, denominator mismatch, ratio reversal, count-prevalence confusion and silent missing-case exclusion. Finish revision by reproducing one analysis from raw frequencies through a checked table and a bounded policy sentence without relying on automatic output labels.
Create the same two-way table with counts, row percentages and column percentages, then write the exact question each display answers. Reverse a relative-risk reference group and explain the changed number without changing the underlying association. Add a third variable and compare stratum-specific tables, watching for sparse cells and confounding rather than searching for a preferred result.
Test two recodes of an ordered category and record what information the binary split removes. Use an unweighted count alongside a weighted percentage and label why both appear. Audit table totals after every filter and retain a simple flow of excluded cases. For reporting practice, write a headline that states the group comparison accurately, then a method note that lets another analyst reproduce it.
End with the strongest interpretation supported by a descriptive table and the design needed for a causal claim.
Working through Tabulation and Data Analysis in SOST10021? Sia is AskSia’s AI Statistics tutor — ask any SOST10021 Tabulation and Data Analysis question and get a clear, step-by-step explanation grounded in how SOST10021 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.