Series ID: EDKS-ADV-ZH-0019 · How to Learn Advanced English (Chinese Edition) · Lesson No.019 · C1 → C2
Write Results That Reveal the Pattern Without Arguing Ahead of the Evidence | 把研究结果写清楚，而不是提前替证据下结论
A Results section should make the evidence easy to see without making the reader accept an explanation the data have not yet earned.
Results 的任务，是让证据模式变得清楚，而不是在读者还没看清证据之前，就替证据把解释、机制和结论全部说完。
This is harder than it sounds. Raw data are not yet Results. But once you select, summarise, order and highlight the data, you have already begun to shape what the reader sees.
所以，Results 并不是“完全没有 interpretation”。真正的高级控制是：
enough analysis to reveal the pattern; not so much interpretation that Results turns into Discussion.
足够的分析，让 pattern 可见；不过度解释，避免 Results 偷偷变成 Discussion。
Before we begin: Lesson 015 owns section boundaries; Lesson 019 owns Results depth | 第015课管 section architecture；第019课深入 Results
Lesson 015 established the broad architecture: Methods = what was done; Results = what was found; Discussion = what the findings may mean.
Lesson 018 owns deep Methods reporting.
Lesson 019 asks the next question:
How do you convert collected evidence into a clear Results section without hiding inconvenient findings or arguing beyond the evidence?
1. Manchester: report results systematically and highlight what matters | Results 要系统呈现，也要 highlight
The University of Manchester Academic Phrasebank describes Results sections as systematic, detailed presentations of findings. In quantitative work, tables and figures often carry evidence while prose locates and highlights the relevant pattern; in qualitative work, writers identify themes and illustrate them with excerpts or other source material. More elaborate commentary is normally reserved for Discussion, although some research articles combine Results and Discussion.
University of Manchester Academic Phrasebank | Reporting Results
2. Elsevier: include negative results and do not duplicate tables | Results 不是成功故事
Elsevier’s Results guidance recommends reporting multiple hypotheses separately, including negative results where relevant, using tables/figures to communicate patterns, and referring to visual displays without repeating all the information already shown.
Elsevier | How to Write the Results Section
3. Nature: statistical information must be inspectable | 统计结果要可检查
Nature’s current formatting guidance requires comprehensive statistical information in research papers, including exact sample sizes where relevant, definitions of error bars, and appropriate reporting of statistical tests and probability values.
Nature | Initial Submission: Statistical Information
4. Results begins with the research question, not the prettiest number | Results 第一层顺序由 question 决定
If your primary question is:
Does structured feedback improve one-week independent revision?
then the first major result should answer that question.
Do not open with:
Participants liked the feedback.
unless preference is the primary outcome.
5. Build a question → outcome → result map | 建立 research lineage
| Research question | Outcome / evidence | Result location |
|---|---|---|
| Does X improve Y? | primary outcome Y | Results section 1 |
| Does effect persist? | delayed outcome | Results section 2 |
| How did participants experience X? | interview themes | Results section 3 |
6. Primary outcomes deserve priority | Primary outcome 不能被 secondary result 抢走
Primary outcomes are not merely the most interesting findings after analysis. They are the outcomes defined as central to the study’s main test.
7. Do not promote a positive secondary result when the primary result is null | 这是常见 spin
Primary outcome:
No clear improvement in independent writing.
Secondary outcome:
Higher self-reported confidence.
Bad Results emphasis:
The intervention improved learner outcomes.
This changes the study.
8. Make the primary result easy to locate | Reader 不应该找五分钟才知道主结果
Use:
- subheading;
- opening sentence;
- main table/figure;
- clear numerical estimate.
9. Results is not raw data | Data ≠ Result
Raw data:
61, 64, 72, 70, 69…
Result:
Mean score was 68.4 (SD 6.1).
Pattern:
Scores were higher in the structured-feedback group than in the comparison group.
10. Results transforms data into an inspectable pattern | Pattern selection 是必要分析
You have to choose:
- summary measure;
- comparison;
- time point;
- visualisation;
- ordering.
That is analytical work, but not yet full interpretation.
11. Observation and explanation need different grammatical homes | Fact 与 explanation 分开
Result:
Mean response time fell by 18%.
Discussion:
The reduction may reflect the simplified workflow.
12. Do not make causal claims in Results unless the design and sentence genuinely support them | Results 不是 causation shortcut
Weak:
The intervention improved memory because learners retrieved the material more deeply.
Better Results:
The intervention group scored 7.4 points higher on the delayed test.
13. Causal effect estimates can still appear in Results | 如果 design 合适，结果可以是 causal estimate
Example:
Random assignment to the intervention increased the mean delayed-test score by 5.2 points.
But mechanism interpretation remains a Discussion job.
14. “Improved” can be legitimate or inflated depending on design | 动词强度取决于研究设计
Randomised comparison:
The intervention improved the measured outcome.
Cross-sectional association:
Intervention use was associated with the measured outcome.
15. Lesson 014 controls certainty here | Results 也需要 certainty calibration
Lesson 014 | Calibrate Certainty Without Becoming Vague
16. Directly measured facts usually need less hedging | 直接测到的状态可以直接写
Forty-eight participants completed the follow-up.
Not:
Approximately 48 participants may have completed…
17. Estimated effects need estimate + uncertainty | Estimate 最好成对出现
Example:
The adjusted mean difference was 4.2 points (95% CI 1.1 to 7.3).
One sentence communicates:
- direction;
- magnitude;
- precision.
18. Do not report p-value without effect | p-value 不是完整 Result
Weak:
There was a significant difference (p = .03).
Better:
The intervention group scored 4.2 points higher (95% CI 1.1–7.3; p = .03).
19. Exact reporting conventions depend on field | 不同 discipline 的统计格式不同
Follow:
- target journal;
- discipline style;
- reporting guideline;
- statistical method.
20. Statistical significance and practical importance are different | “显著”有两个世界
A tiny effect can be statistically precise.
A large-looking effect can be imprecise.
Results should report the estimate; Discussion evaluates importance.
21. Avoid “highly significant” as rhetorical emphasis | highly significant 不是“非常重要”
If p-value is small, report it according to field conventions.
Do not make statistical language do rhetorical work.
22. Null results need the same visibility as positive results | Null result 不是 failure
Weak:
No effect was found.
Better:
The estimated difference at six months was 0.8 points (95% CI −2.1 to 3.7), providing no clear evidence of a persistent between-group difference.
23. “No significant difference” is not equivalence | 再次强调：non-significant ≠ same
A wide interval can include both benefit and harm.
24. Equivalence/non-inferiority needs the right design | “差不多”必须被正式测试
Do not infer equivalence from ordinary null-hypothesis testing.
25. Negative results should be reported where relevant | Elsevier 明确强调 negative results
If the study tested three hypotheses and one failed, report it.
Do not turn Results into a highlight reel.
26. Unexpected findings deserve transparent status | Unexpected ≠ embarrassing
State the observed pattern.
Do not rush to explain it in Results.
27. “Contrary to expectations” can be enough | 把 surprise 标出来，不必当场解释
Contrary to the preregistered prediction, the groups did not differ on delayed accuracy.
28. Exploratory findings need labels | Exploratory 结果不要伪装成 confirmatory
In an exploratory analysis…
A post-hoc subgroup analysis indicated…
29. Labels protect inferential status | Label 不是自我贬低
Exploratory results can be valuable.
The label tells the reader how they entered the analysis.
30. Preregistered hypotheses should remain distinguishable | Confirmatory / exploratory 时间顺序要可见
31. Order results by research logic | Results 顺序不是按 software output
Possible ordering:
- sample/flow;
- primary outcome;
- secondary outcomes;
- subgroups;
- exploratory analyses.
32. Participant flow can be a Result | Recruitment / retention 最终数量属于 Result
Methods says planned recruitment/eligibility.
Results says:
132 were screened, 104 enrolled, 96 completed the primary outcome.
33. Denominator changes must be visible | n 变了，读者必须知道
Table 1 n = 100.
Table 2 n = 84.
Why?
34. Exact n protects against silent attrition | Nature guidance emphasises exact sample sizes where relevant
Do not write approximately 100 when exact analysis n is 87.
35. Baseline/descriptive characteristics have a job | 描述样本，不是证明 randomisation “成功”
Use descriptive information to show who/what was analysed.
36. Avoid significance testing of baseline randomisation unless methodologically justified | Randomisation 已由设计产生，不靠 baseline p-value 证明
Follow field guidance.
37. Describe missingness as a finding | Missing data 也有 Results component
Methods:
how missing data would be handled.
Results:
how much data were actually missing and where.
38. Report exclusions transparently | 谁被排除，多少，为什么
If participant flow matters, use a flow diagram.
39. Tables should carry dense evidence efficiently | Table 是 high-density evidence surface
Use tables when readers need:
- exact values;
- several outcomes;
- multiple groups/time points;
- model coefficients;
- participant characteristics.
40. Figures should reveal shape/pattern | Figure 擅长关系与趋势
Use figures for:
- time trends;
- distributions;
- dose-response relationships;
- interactions;
- uncertainty around estimates;
- spatial or network patterns.
41. Choose table or figure by reader task | 不要因为软件容易画图就画图
Exact comparison?
Pattern recognition?
Distribution?
Trend?
42. Do not duplicate table as figure | 同一信息不要两次占空间
43. Do not repeat every table cell in prose | Manchester / Elsevier 都强调 highlight，而不是复读
Weak:
Group A was 61 at T1, 66 at T2, 68 at T3. Group B was 60 at T1, 62 at T2, 63 at T3.
If table already shows this:
Group differences widened over time, with the largest separation at T3.
44. Highlighting is not overinterpretation | Highlight 可以指出 pattern
The largest difference occurred at delayed follow-up.
This is still descriptive.
45. “Why” usually crosses into Discussion | 一旦回答 why，就要警觉
The largest difference occurred at delayed follow-up because retrieval strengthened memory traces.
The second clause is Discussion.
46. Use location statements for figures/tables | 先告诉读者看哪里
Figure 2 shows the change in revision score across the three time points.
47. Then use highlighting statements | 再告诉读者注意什么
The groups were similar immediately after training but diverged at one-week follow-up.
48. Table titles should identify content | Table 2. Adjusted differences in…
Follow venue formatting.
49. Figure legends should define uncertainty and symbols | Error bars 到底是什么？
SD?
SE?
95% CI?
Credible interval?
50. Error bars without definition are ambiguous | Nature explicitly requires definition of error bars
51. Axes can lie without false numbers | Visual rhetoric 也会 distort
Truncated axes.
Unequal intervals.
Log scales without explanation.
Overplotting.
52. Scale should support accurate perception | 不要把小差异画成悬崖
53. Bar charts can hide distributions | 平均值相同，分布可能完全不同
When distribution matters, consider:
- dot plots;
- box plots;
- violin plots;
- histograms;
- raw-data overlays.
Follow field norms and accessibility.
54. Avoid decorative chart complexity | Nature advises figures be as simple as compatible with clarity
55. Every figure should answer a question | Figure 不应该只是“看起来专业”
56. Statistical tables need understandable labels | Model 1 / Model 2 要说明区别
57. Coefficients need units/context | 一个 beta = 0.42 不够
What variable?
What unit?
Adjusted for what?
58. Relative effects need absolute context where important | RR/OR 不应完全取代 baseline risk
59. Percentage change needs starting value | +100% can mean 1 → 2
60. Means need spread | Mean without SD/SE/CI may hide variability
61. Medians often need IQR/range | 尤其 skewed data
62. Sample size belongs beside estimates | n 不应该藏在 Methods only
63. Confidence intervals communicate precision | CI 不只是“有没有跨 0”
Read:
- direction;
- range;
- plausible magnitude;
- precision.
64. Do not interpret CI as a deterministic probability statement unless the framework supports it | 统计语言要跟方法一致
65. Bayesian results need Bayesian reporting | Credible intervals / posterior probabilities 等不要翻成频率学派术语
66. Qualitative Results reveal themes, patterns and variation | Qualitative Results 不是“放几个 quote”
Manchester notes that qualitative reporting often highlights themes emerging from analysis and illustrates them with excerpts from raw data.
67. Theme needs an analytical proposition | Theme 名字不能只是 topic
Weak theme:
Feedback.
Stronger:
Feedback became useful when it identified a next action rather than only an error.
68. Quote illustrates; it does not do the analysis for you | 引文不是分析
Pattern first.
Evidence second.
69. Use representative excerpts | 不要只选最漂亮 quote
Include variation when it changes the theme.
70. Negative/deviant cases matter | 反例可以 refine theme
If one subgroup consistently reports the opposite experience, do not delete it.
71. Frequency language in qualitative Results needs methodological fit | many / most / several 不能乱用
Some qualitative approaches treat frequency as informative.
Others focus on meaning, process or variation rather than counting.
72. Do not quantify qualitative data casually | 5/8 participants said… can be useful or misleading depending design
73. Qualitative Results can contain interpretation | 主题本身已经是分析产物
The distinction is not “no interpretation.”
The distinction is:
analysis grounded in the dataset versus broader explanation/theoretical implication.
74. Mixed-methods Results must preserve evidence type | Quant + Qual 不要被揉成一句 fake certainty
Quant:
scores improved.
Qual:
participants described clearer next actions.
Discussion can integrate mechanism more fully.
75. Joint displays can support mixed-methods integration | Joint display 把 strands 并排，而不是混掉
| Quantitative pattern | Qualitative pattern | Relation |
|---|---|---|
| higher revision score | clearer next-action reports | convergent |
| no accuracy gain | few comments on grammar | compatible |
76. Relation labels belong cautiously in Results | Convergent / divergent 可以描述 evidence relation
Mechanistic explanation belongs later.
77. Computational Results need benchmark transparency | Model performance 不是一个 accuracy 数字
Report:
- dataset split;
- benchmark;
- metric;
- uncertainty/variation;
- baseline model;
- number of runs where relevant.
78. Best run is not average performance | 不要只报最漂亮 seed
Where stochasticity matters, report distribution/mean across runs according to field norms.
79. Test-set contamination invalidates apparent performance | 如果 benchmark 泄漏，Result interpretation 变了
Methods should describe prevention; Results should not hide known contamination.
80. Ablation results can clarify component contribution | Ablation 是 Result，mechanism interpretation later
81. Benchmark cherry-picking is selective reporting | 只选模型擅长的数据集会制造假 superiority
82. Clinical/health Results often need participant flow + harms | Design-specific guidelines matter
CONSORT/STROBE/PRISMA and other EQUATOR guidelines specify study-type reporting needs.
EQUATOR Network | Reporting Guidelines
83. Harms/adverse events belong alongside benefits | 不能只报 efficacy
84. Systematic-review Results need study flow | How many records → included studies?
85. Synthesis Results need heterogeneity/uncertainty | Meta-analysis 不只是 pooled number
86. Risk-of-bias results affect evidence interpretation | Review Results 可以报告 risk-of-bias pattern
87. Narrative synthesis still needs explicit pattern | “Studies were mixed” 不是 adequate synthesis
Say:
Short-term effects were consistently positive, while delayed outcomes varied by study duration and baseline proficiency.
88. Results can be organised by hypothesis | 多 hypothesis 时分开写
Elsevier recommends separate description to avoid confusion.
89. Or by research question | Often clearer for readers
90. Or by outcome hierarchy | Primary → secondary → exploratory
91. Or by analytic sequence | Descriptive → primary model → sensitivity
92. Do not use software-output order automatically | SPSS/R table order 不是 reader logic
93. Sensitivity analyses are Results | Methods defines; Results reports
The primary estimate remained similar after excluding…
94. Robustness should say what survived | “Results were robust” 太空
The direction and approximate magnitude were unchanged across three alternative specifications.
95. Subgroup analyses need special caution | Subgroup 是 false-positive 高风险区
State:
- pre-specified or exploratory;
- interaction evidence;
- sample size/precision.
96. “Significant in group A but not group B” does not prove groups differ | Compare interaction directly
97. Avoid dichotomising continuous evidence | p .05 不是 cliff
Report estimates/uncertainty.
98. Multiplicity matters | 大量 subgroup tests 会制造 chance findings
Methods sets correction strategy; Results reports accordingly.
99. Secondary analyses should not crowd out primary answer | Results 需要 hierarchy
100. Length should follow information value | Results 不是所有 output 的 archive
Supplementary material can carry:
- secondary tables;
- diagnostic plots;
- full model output;
- additional robustness checks.
Main text preserves the inferential spine.
101. Supplement should not hide inconvenient primary evidence | Main result 不能被“移到 supplement”消失
102. Results should not contain literature review paragraphs | 不要在每个 result 后重新讲前人研究
That is mainly Discussion.
103. Limited comparison language may appear in combined sections | If venue combines Results & Discussion
Then signal moves explicitly.
104. Combined Results & Discussion needs move-level clarity | Result → Interpretation → Comparison → Limitation
Manchester notes this combined form is common in some research articles.
105. Do not impose a false firewall | Results can include minimal commentary
The correct boundary is not “zero interpretation.”
The correct boundary is:
Do not claim more than the displayed evidence directly supports before broader explanation is considered.
106. Words like “interesting” and “surprising” are usually unnecessary | Results 不需要作者情绪
Interestingly…
Ask whether the pattern itself can be stated.
107. “Notably” can signal emphasis but should not replace hierarchy | Why is it notable?
108. Avoid “clearly” | 数据应该自己显示 clarity
109. Avoid “remarkably” | 除非是 defined comparison, not rhetoric
110. Avoid “as expected” if it encourages confirmation bias | 预期与否可以写，但不要淡化 unexpected results
111. Use neutral highlighting | Neutral ≠ dull
The largest difference occurred…
Scores remained stable…
No clear association was detected…
112. Describe trend direction precisely | increased / decreased / plateaued / fluctuated / converged / diverged
113. Describe magnitude carefully | small / moderate / large only if defined or contextualised
114. “Substantial” needs a benchmark | Compared with what?
115. “Meaningful” usually crosses toward Discussion unless criterion pre-defined | Practical meaning needs context
116. Use absolute units where possible | 6.1 points more transparent than “substantially higher”
117. Relative change can mislead | +50% from 2 to 3
118. Baseline denominators protect interpretation | Risk 1% → 2% = doubled but +1 percentage point
119. Report both relative and absolute effects when decision-relevant | Depending on field
120. Do not hide raw scale | Standardised effects can help comparison but readers may also need original units
121. Use meaningful decimal precision | 74.23871 often false precision
122. Match decimals to measurement precision | Instrument cannot justify six decimals
123. Percentages with tiny n can mislead | 75% when n=4
Give counts.
124. Sample composition can be a Result | Who completed study?
Relevant demographic/context characteristics.
125. But demographic table is not a biography | Include characteristics relevant to interpretation/reporting norms
126. Baseline imbalance can be reported descriptively | Discussion/analysis decides consequences
127. Protocol deviations can have Results manifestations | How many received planned intervention?
128. Adherence/dose can be Result | Intervention delivered vs received
129. Manipulation checks can be Result | Did manipulation change intended proximal state?
130. But manipulation check is not the primary outcome unless designed so | hierarchy matters
131. Reliability estimates can appear in Results | If estimated in current sample
132. Inter-rater agreement can be Result | Methods says process; Results gives observed agreement
133. Model diagnostics can be Results or supplement | Depends on importance/venue
134. Assumption failure may change main analysis | If so, make visible
135. Data-quality checks can be Result | Especially if they affect usable n
136. Do not report every check if it does not affect inference | main text economy
137. Results section openings should orient | Start with analysis/question
We first examined whether…
To test the primary hypothesis…
138. Then report the estimate | estimate before commentary
139. Then add necessary pattern highlight | one or two sentences
140. Then move to next question | Avoid mini Discussion after every number
141. Paragraph architecture | Results paragraph model
QUESTION/ANALYSIS → ESTIMATE → UNCERTAINTY → PATTERN HIGHLIGHT → SECONDARY DETAIL.
142. Example quantitative paragraph | 示例
To test the primary hypothesis, we compared one-week independent revision scores between groups. The structured-feedback group scored a mean of 74.2 (SD 7.8), compared with 68.1 (SD 8.3) in the standard-feedback group. The adjusted mean difference was 6.0 points (95% CI 2.8–9.2). The largest group difference occurred in evidence integration, whereas sentence-level accuracy differed little between conditions.
143. Why this works | 它没有解释 mechanism
It reports:
- question;
- descriptives;
- effect estimate;
- precision;
- subdimension pattern.
144. Discussion can then ask why | 为什么 evidence integration 更受影响？
145. Example null paragraph | Null result 示例
At 12-week follow-up, mean scores were 72.8 and 71.9 in the structured and standard groups, respectively. The adjusted between-group difference was 0.9 points (95% CI −2.4 to 4.2). Thus, the follow-up data did not show a clear persistent group difference.
146. Note the wording | “did not show a clear difference” 不等于“proved no effect”
147. Example exploratory subgroup paragraph | Exploratory 示例
In a post-hoc analysis, the estimated one-week difference was larger among learners with lower baseline revision scores than among those with higher baseline scores. Because this subgroup analysis was not preregistered and estimates were imprecise, it is reported as exploratory.
148. Minimal interpretation can include inferential status | “reported as exploratory” 属于 evidence status, not mechanism
149. Example qualitative theme paragraph | Qualitative 示例
Participants described feedback as most useful when it identified a specific next action. This theme appeared across both higher- and lower-scoring participants. One learner explained that comments became easier to use when they stated “what to change next” rather than only identifying a weakness. A smaller set of participants, however, reported that highly structured prompts felt restrictive when they already knew how they wanted to revise.
150. What this qualitative paragraph does | Theme + evidence + variation
Broader explanation of why structure helps or constrains belongs in Discussion.
151. Example mixed-methods joint result | Mixed Methods 示例
The structured group showed higher evidence-integration scores, while interviews from the same group more frequently described feedback as providing a clear revision sequence. Sentence-level accuracy showed little group difference, and grammar-related interview comments were uncommon.
152. This reports convergence without claiming mechanism | “clear revision sequence caused the score gain” would go further
153. The selective-reporting audit | Selective reporting 审计
List every:
- pre-specified outcome;
- time point;
- group;
- planned analysis.
Where is each reported?
154. The primary-outcome audit | Primary Result 能否在 10 秒内找到？
155. The null-result audit | Null/negative finding 是否被缩到一句 footnote？
156. The denominator audit | 每个百分比/estimate 的 n 是否清楚？
157. The attrition audit | 各 time point 的 n 是否变化？
158. The estimate audit | 报 effect size / magnitude，而不只 p-value？
159. The precision audit | uncertainty 是否可见？
160. The causality audit | 动词是否超过 design？
161. The table-duplication audit | 正文是否在逐格复读？
162. The figure-scale audit | Visual scale 是否放大/缩小差异？
163. The exploratory-label audit | Post-hoc 是否被标记？
164. The subgroup audit | 是否把 “A significant, B not significant” 误写成 subgroup difference？
165. The primary-vs-secondary audit | Positive secondary 是否抢走 null primary？
166. The missingness audit | Missing data 的实际数量是否报告？
167. The orphan-result audit | 每个 Result 能追溯到 Method 吗？
168. The Discussion-leakage audit | Results 里有多少 because / suggests mechanism / should？
Highlight:
- because;
- therefore this means;
- may be due to;
- supports the theory that;
- should be implemented.
Inspect whether they belong later.
169. The rhetoric audit | clearly / remarkably / importantly / surprisingly 是否真正必要？
170. The result-order audit | 排序是否跟 research question，而不是 software output？
171. The visual-accessibility audit | 图表是否可读、标签完整、颜色依赖是否安全？
172. The reader reconstruction test | 只给 Results，读者能回答什么？
They should be able to reconstruct:
- sample actually analysed;
- primary pattern;
- magnitude;
- uncertainty;
- null/negative findings;
- exploratory status.
They should not be forced to accept an untested mechanism.
173. Worked case: full fictional Results set | 完整虚构案例
All values below are fictional teaching material.
Study:
96 advanced bilingual learners randomised to structured action-oriented feedback or standard evaluative feedback.
Primary:
one-week independent revision.
Secondary:
evidence integration, sentence accuracy, self-reported feedback usability.
Delayed:
12-week revision task.
174. Participant flow | 样本流
Of 124 learners assessed for eligibility, 100 enrolled and 96 completed the one-week primary outcome (48 per group). At 12 weeks, follow-up data were available for 82 learners (42 structured; 40 standard).
175. Primary outcome | 主结果
At one week, mean independent-revision scores were 74.2 (SD 7.8) in the structured group and 68.1 (SD 8.3) in the standard group. The adjusted mean difference was 6.0 points (95% CI 2.8–9.2).
176. Secondary outcome | Evidence integration
The largest rubric difference was observed for evidence integration (mean difference 1.4 points on the 0–5 scale, 95% CI 0.7–2.1).
177. Secondary outcome | Sentence accuracy
Sentence-level accuracy differed by 0.2 points (95% CI −0.3 to 0.7), with no clear group separation.
178. Usability | Self-report
Structured-feedback participants rated feedback usability higher on the 1–7 scale (5.9 vs 4.8).
179. Delayed outcome | 12 weeks
At 12 weeks, mean revision scores were 72.8 and 71.9. The adjusted difference was 0.9 points (95% CI −2.4 to 4.2).
180. Missingness | Follow-up
Fourteen participants lacked 12-week outcome data; attrition was slightly higher in the standard group.
181. Sensitivity analysis | Missing data
Multiple-imputation sensitivity analysis produced a similar 12-week estimate and did not materially change the conclusion that the delayed group difference was imprecise.
182. Exploratory subgroup | Baseline proficiency
In a post-hoc analysis, the one-week group difference was larger among learners below the sample median at baseline. The group-by-baseline interaction estimate was imprecise and is reported as exploratory.
183. What the Results section still does not say | Results 不应该说什么？
It does not yet say:
- structured prompts caused deeper metacognition;
- the intervention failed long-term;
- advanced learners need less support;
- schools should adopt the method.
Those require Discussion-level reasoning.
184. A possible Results narrative | 把数字组织成 pattern
Structured feedback was associated with a clear one-week advantage on the primary independent-revision outcome, with the largest difference in evidence integration and little difference in sentence-level accuracy. The between-group difference was no longer clear at 12 weeks, where the estimate was small and imprecise. Follow-up attrition reduced the delayed-analysis sample, although sensitivity analysis produced a similar estimate. An exploratory analysis suggested a larger short-term difference among lower-baseline learners, but the interaction estimate was imprecise.
185. Is that “too interpretive”? | 注意 sentence strength
The paragraph summarises observed evidence status.
It does not offer a mechanism.
That is acceptable Results-level synthesis in many genres.
186. Practice A: result vs discussion | 练习 A
Classify:
The intervention group scored 8% higher.
The gain likely occurred because feedback reduced cognitive load.
187. Practice B: repair p-only result | 练习 B
The difference was significant (p = .02).
Rewrite with magnitude.
188. Practice C: null result | 练习 C
Estimate = 1.0 point; 95% CI −3 to 5.
Write one calibrated result sentence.
189. Practice D: table highlight | 练习 D
Table has 12 values.
The main pattern: groups diverge only at final time point.
Write one prose sentence.
190. Practice E: selective reporting | 练习 E
Primary outcome null; secondary confidence positive.
Write a fair two-sentence Results summary.
191. Practice F: subgroup trap | 练习 F
Effect significant in women, non-significant in men.
What analysis is needed before claiming subgroup difference?
192. Practice G: percentage denominator | 练习 G
75% preferred A; n = 12.
Report transparently.
193. Practice H: qualitative theme | 练习 H
Create theme + quote + deviant case.
194. Practice I: exploratory label | 练习 I
Subgroup found after inspecting data.
Write one sentence.
195. Practice J: figure caption | 练习 J
Write a caption defining groups, outcome and 95% CI error bars.
196. Practice K: mixed methods | 练习 K
Quant gain + interview theme.
Write a Results sentence that reports convergence without claiming cause.
197. Model answers | 示范答案
A: First = Results. Second = Discussion.
B: The intervention group scored 5.3 points higher than the comparison group (95% CI 0.9–9.7; p = .02).
C: The estimated difference was 1.0 point (95% CI −3.0 to 5.0), providing no clear evidence of a between-group difference and remaining compatible with effects in either direction.
D: The groups were similar at the first two assessments but diverged at the final time point.
E: The primary performance outcome did not differ clearly between groups. Participants in the intervention group nevertheless reported higher confidence, a secondary outcome.
F: Test the group-by-sex interaction directly; separate significance tests do not establish a difference in effects.
G: Nine of 12 participants (75%) preferred Option A.
I: In a post-hoc exploratory analysis, the estimated effect was larger among…; this subgroup pattern was not prespecified.
K: The group with higher revision scores also more frequently described feedback as giving a clear next action; these patterns converged across the quantitative and interview strands.
198. The 20-minute Results drill | 20 分钟训练
| Time | Task |
|---|---|
| 3 min | map research question → outcome |
| 4 min | report estimate + uncertainty |
| 4 min | write table/figure highlight |
| 4 min | report null/exploratory result |
| 5 min | discussion-leakage audit |
199. The 45-minute growth session | 45 分钟增长模式
| Time | Task |
|---|---|
| 8 min | analyse one target-journal Results section |
| 8 min | build outcome hierarchy |
| 10 min | draft quantitative/qualitative Results |
| 10 min | table/figure commentary |
| 9 min | selective-reporting + certainty audit |
200. The 90-minute deep session | 90 分钟深度模式
| Time | Task |
|---|---|
| 15 min | research-question/result lineage map |
| 15 min | participant flow/descriptives |
| 20 min | primary/secondary Results draft |
| 15 min | tables/figures |
| 10 min | null/exploratory/sensitivity reporting |
| 15 min | anti-spin + Discussion leakage audit |
201. Seven-day Results cycle | 七天训练循环
- Day 1: outcome hierarchy and participant flow.
- Day 2: descriptive statistics and sample reporting.
- Day 3: effect estimates and uncertainty.
- Day 4: tables/figures and highlighting.
- Day 5: null/negative/exploratory findings.
- Day 6: qualitative/mixed-methods reporting.
- Day 7: full Results benchmark.
202. Twelve-week C1–C2 Results progression | 12 周路线
| Weeks | Focus | Output |
|---|---|---|
| 1–2 | data → result → pattern | result sentences |
| 3–4 | effect estimates + uncertainty | quantitative paragraphs |
| 5–6 | tables/figures + visual integrity | evidence displays |
| 7–8 | null, negative, exploratory, subgroup | balanced Results sections |
| 9–10 | qualitative + mixed methods | theme/joint Results |
| 11–12 | discipline-specific transfer | submission-ready Results portfolio |
203. Monthly benchmark | 每月基准任务
Use a permitted or fictional dataset.
Produce:
- research-question → outcome map;
- participant-flow paragraph;
- descriptive table;
- primary outcome paragraph;
- secondary outcome paragraph;
- null/negative result paragraph;
- exploratory result paragraph;
- one figure;
- one figure-highlight paragraph;
- sensitivity-analysis sentence;
- selective-reporting audit;
- denominator audit;
- Discussion-leakage audit;
- primary-vs-secondary audit;
- Method → Result lineage map.
204. First weak-link diagnosis | 第一个薄弱环节
| Symptom | Likely weak link | Repair |
|---|---|---|
| Results starts with secondary finding | outcome hierarchy lost | primary question first |
| p-values everywhere, no magnitude | estimate reporting weak | effect + uncertainty |
| null result says “no effect” | equivalence confusion | report estimate/CI |
| table repeated in prose | highlighting skill weak | state pattern only |
| positive secondary dominates | selective reporting/spin | restore primary outcome |
| subgroup claim from separate p-values | interaction reasoning weak | test subgroup difference directly |
| qualitative Results = quote list | theme analysis missing | theme → evidence → variation |
| Results explains why | Discussion leakage | move mechanism later |
205. What not to do | 不要这样写 Results
- Do not report only findings that support the hypothesis.
- Do not let positive secondary outcomes replace a null primary outcome.
- Do not report p-values without effect magnitude where the field expects estimates.
- Do not equate non-significance with equivalence.
- Do not repeat every table or figure value in prose.
- Do not make causal/mechanistic explanations before the evidence has been interpreted.
- Do not hide denominator changes, missing data or attrition.
- Do not present exploratory analyses as preregistered hypotheses.
- Do not use visual scales that exaggerate or conceal differences.
- Do not confuse a clear Results narrative with a persuasive sales narrative.
206. Research and reference floor | 研究与参考基础
- University of Manchester Academic Phrasebank | Reporting Results — systematic result presentation, table/figure location and highlighting, positive/negative findings and Results–Discussion boundaries.
- Elsevier | How to Write the Results Section — reporting multiple hypotheses, negative findings, tables/figures and avoiding unnecessary duplication.
- Nature | Initial Submission — statistical reporting, exact sample sizes, repeated measurements and error-bar definitions.
- Nature | Formatting Guide — figure simplicity, legends, statistical definitions and display-item integrity.
- EQUATOR Network | Reporting Guidelines — design-specific reporting requirements for trials, observational studies, reviews, qualitative studies and other designs.
- Academic Phrasebank | Being Cautious — calibrating inferential language.
207. Canonical eduKate routes | eduKate 主页面路由
- Lesson 014 | Calibrate Certainty Without Becoming Vague
- Lesson 015 | Methods, Results and Discussion
- Lesson 016 | Abstract Compression
- Lesson 018 | Methods for Reproducibility
208. SEO language map | 本课自然覆盖的搜索意图
This lesson naturally serves learners searching for how to write results section, reporting research results, Results section examples, research paper results, quantitative results writing, qualitative results writing, effect estimates, confidence intervals, null findings, negative results, tables and figures academic writing, exploratory analysis reporting, subgroup analysis, research spin, C1 academic writing, C2 academic English, Results 怎么写, 研究结果怎么写, null result 怎么写, confidence interval 怎么报告, 表格图表学术写作, 定量结果, 定性结果, 中文母语学术英语 and C1 C2 research writing.
209. The one-page Results operating system | 一页 Results 操作系统
- Return to the research question.
- Put the primary outcome first.
- Report participant flow and actual n where relevant.
- Report direction + magnitude + uncertainty.
- Do not rely on p-values alone.
- Report null/negative findings honestly.
- Keep primary and secondary outcomes distinct.
- Label exploratory/post-hoc analyses.
- Use tables for dense exact values.
- Use figures for shape/trend/distribution.
- Highlight the pattern; do not repeat every cell.
- Define error bars and denominators.
- Preserve missingness/attrition.
- For qualitative work: theme → evidence → variation.
- For mixed methods: preserve evidence types before integration.
- Run selective-reporting and spin audits.
- Run Discussion-leakage audit.
- Make every Result traceable to a Method.
210. Final assignment | 最终作业
Choose a permitted or fictional dataset with at least one primary outcome and one secondary or exploratory analysis.
Complete this chain:
- Write the research question.
- Map primary/secondary/exploratory outcomes.
- Write participant flow.
- Create one descriptive table.
- Create one main figure or effect table.
- Write the primary Results paragraph with estimate and uncertainty.
- Write one secondary result.
- Write one null/negative result honestly.
- Write one exploratory result with a clear label.
- Report actual denominators/sample sizes.
- Report missingness/attrition.
- Write one sensitivity-analysis sentence.
- If qualitative, write one theme paragraph with supporting evidence and a deviant case.
- If mixed methods, create one joint display.
- Run the primary-outcome audit.
- Run the p-value-only audit.
- Run the selective-reporting audit.
- Run the subgroup audit.
- Run the table-duplication audit.
- Run the Discussion-leakage audit.
Then ask:
If a reader disagrees with my interpretation later, would they still be able to see clearly what the evidence itself showed here?
如果读者后来不同意我的 Discussion，他们仍然能不能只看 Results，就清楚知道证据本身显示了什么？
If yes, your Results section is doing its job.
If no, you may have mixed evidence and argument too early.
Continue | 继续
The next lesson will complete the research-paper core by going deep into Discussion: how to interpret findings, compare rival explanations, connect to prior research, handle limitations and implications, and finish with a conclusion no stronger than the evidence.
下一课进入 Discussion 深层写作：怎样解释 finding、比较替代解释、连接文献、处理 limitation 与 implication，并让最终结论不超过 evidence 真正支持的范围。
Next: EDKS-ADV-ZH-0020 · Lesson No.020 · Write a Discussion That Explains the Findings Without Explaining Them Away · 解释研究结果，但不要把证据解释没了
