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 strong Results section makes the evidence easy to see before the writer tells the reader what to think about it.

强 Results section 的第一任务，是先让 evidence 本身清楚可见；不要在 reader 还没看见 pattern 之前，就急着替 evidence 下结论。

This does not mean Results is a warehouse of raw data. Data become results only after the writer organises them around the research question, identifies the important pattern and reports it clearly. But the Results section should normally stop before it becomes a full explanation of mechanism, literature comparison or broad implication.

因此，Results 的核心边界是：

data → relevant pattern → transparent result.

而不是：

data → preferred story → final argument.

Before we begin: Lesson 015 owns section architecture; Lesson 019 owns Results depth | 第015课管 section architecture；第019课深入 Results

Lesson 015 established the broad Methods–Results–Discussion separation.

Lesson 018 owns deep Methods transparency and reproducibility.

Lesson 019 owns the next handoff: how to turn analysed data into visible findings without smuggling the Discussion into Results.

1. Manchester: Results should present findings systematically and in detail | Results 首先要 systematic + detailed

The University of Manchester Academic Phrasebank describes the Results section as the place where findings are presented systematically and in detail. Quantitative papers often use tables and figures plus highlighting statements; qualitative papers identify and illustrate themes. More elaborate commentary is usually reserved for Discussion, although some fields combine Results and Discussion.

University of Manchester Academic Phrasebank | Reporting Results

2. Elsevier: primary question first, then secondary findings | 先回答最重要的问题

Elsevier’s current guidance recommends presenting findings systematically, addressing the primary research question first and then secondary questions, including negative results where relevant, and using tables/figures without duplicating their content in prose.

Elsevier | Writing the Results Section

3. EQUATOR: reporting completeness depends on study design | Results 也有 design-specific reporting floor

The EQUATOR Network maintains reporting guidelines across many designs. Outcome reporting expectations differ across randomised trials, observational studies, systematic reviews, qualitative research, diagnostic studies and other genres.

EQUATOR Network | Selecting the Appropriate Reporting Guideline

4. Data are not Results | Data ≠ Results

Raw data:

61, 65, 67, 72, 74…

Result:

Mean score increased from 64.2 to 71.6.

Interpretation:

The intervention may have improved transfer.

These are different levels.

5. Results already contain analytical selection | Results 不是“零 interpretation”

Choosing:

which outcome to report first;
which comparison matters;
which pattern to highlight;
which figure best reveals it;

already requires judgement.

The advanced skill is not eliminating judgement. It is keeping the judgement close to what the data directly support.

6. The Results section should answer the research question visibly | RQ → Result 必须直接

For each research question, readers should be able to locate the corresponding result without reconstructing the entire paper.

7. Build an RQ–Result matrix before drafting | 先建立 RQ–Result matrix

Research question	Primary outcome/result	Secondary result	Table/Figure
Does structured feedback improve independent revision?	between-group revision score	rubric dimensions	Table 2
Does the effect persist?	12-week score	retention by baseline proficiency	Figure 2

8. If a research question has no result, something is wrong | Question 没有 answer = architecture problem

Possibilities:

analysis missing;
measure failed;
question too broad;
result omitted;
study design cannot answer it.

9. If a major result has no research-question parent, inspect it | Orphan Result

It may be:

exploratory;
post-hoc;
secondary;
irrelevant.

Label it correctly.

10. Primary outcome first | Primary outcome 先出现

Do not bury the study’s main test below a collection of attractive secondary findings.

11. Primary outcome is not “the result that worked best” | Primary 不能事后挑

Where the design uses pre-specified outcomes, preserve that hierarchy.

12. Secondary outcomes enrich; they do not replace the primary story | Secondary 负责补充

Useful secondary findings may:

clarify which dimension changed;
identify a trade-off;
show subgroup variation;
suggest a mechanism;
reveal harm.

13. Exploratory analyses must be labelled | Exploratory 不应伪装 confirmatory

Useful language:

In an exploratory analysis…

Post-hoc subgroup analysis indicated…

14. Exploratory does not mean useless | Exploratory 可以产生下一步 hypothesis

But it carries a different inferential status.

15. Report direction, magnitude and precision | 不要只报 “significant”

Minimum useful result often includes:

direction;
magnitude;
uncertainty/precision;
group/time denominator where relevant.

16. “Significant” without an estimate hides the result | 只有 p-value，不够

Weak:

The treatment group performed significantly better.

Stronger:

The treatment group scored 6.1 points higher (95% CI 2.4–9.8).

17. Statistical significance is not effect magnitude | p-value 不能替代 effect size

A tiny effect can be statistically significant in a large sample.

A practically meaningful effect can be imprecisely estimated in a small sample.

18. Practical importance is not a Results adjective | “important effect” 通常属于 Discussion

Results should report the estimate clearly.

Discussion evaluates practical meaning.

19. Confidence intervals are high-information language | CI 可以同时显示 direction + magnitude + precision

Use them where appropriate to the field.

20. Wide intervals deserve visible wording | Precision 不足要可见

The estimated difference was positive but imprecise.

21. Null results are results | Null result 不是空白

Weak:

No effect was found.

Better:

The study did not detect a clear between-group difference at 12 weeks; the estimate was small and imprecise.

22. Non-significance is not equivalence | 不显著 ≠ 相同

Equivalence or non-inferiority claims require appropriate design and analysis.

23. “No evidence of X” is not always “evidence of no X” | absence of evidence ≠ evidence of absence

Precision and power matter.

24. Report negative findings honestly | Negative result 不能因为不好看就删

Elsevier’s Results guidance explicitly recommends reporting negative results when relevant, even when they do not support the hypothesis.

25. Selective outcome reporting creates a distorted study | Selective reporting 改变 study identity

If the Methods declare three outcomes and Results report only one, readers need to know why.

26. Preserve pre-specified hierarchy | Outcome switching 是高风险

Primary null + secondary positive should not become:

The intervention was effective.

27. Adverse findings may be more important than positive outcomes | Harm/trade-off 不能因为篇幅删掉

Examples:

accuracy improves but completion time doubles;
mean benefit with high dropout;
performance improves while error severity worsens.

28. Denominators matter | 百分比没有 denominator 很容易误导

84% completed

is stronger with:

42 of 50 participants (84%) completed.

29. Flow counts matter in longitudinal/trial work | 谁进入、谁完成、谁被分析

Study-specific guidelines may require participant-flow reporting.

30. Attrition should be visible | Attrition 不是 footnote

If 30% disappear before follow-up, the reader needs that state.

31. Missingness belongs in the result state when it affects interpretation | Missing data amount 是 result-adjacent information

Methods explains handling.

Results often report how much data were missing.

32. Baseline description has a job, not a ritual | Baseline table 不是装饰

It helps readers understand the analysed groups/population.

Do not turn it into a p-value hunting exercise unless the design/field specifically requires that logic.

33. Organise Results by research logic | Results 顺序不一定按 analysis 软件输出

Common structures:

RQ order;
hypothesis order;
primary → secondary outcome;
time sequence;
theme sequence;
model sequence.

34. Software output order is not reader order | SPSS/R 输出顺序不是文章结构

35. Use subheadings when they map the evidence | Subheading 要帮助 reader 看问题

Good:

Primary outcome: independent revision

Delayed retention

Exploratory subgroup analyses

36. Avoid result subheadings that already interpret | 不要标题先下结论

Risky:

Structured Feedback Successfully Improves Learning

Better:

Independent Revision Scores

37. Begin each subsection with the question or finding state | 先定位 reader

We first examined whether the groups differed on the primary revision outcome.

38. Then report the result directly | 不要绕三句才说数字

Mean scores were 74.2 and 68.1, respectively.

39. Highlight, do not narrate every calculation | Results 不是 analysis diary

Do not report:

First we ran this command, then inspected this output, then…

That belongs in Methods/analysis documentation.

40. Tables are for structured comparison | Table 擅长让读者查数

Use tables for:

group summaries;
model estimates;
multiple outcomes;
participant characteristics;
categorical counts;

41. Figures are for visible patterns | Figure 擅长趋势、关系、分布

Use figures for:

time trends;
dose-response;
distribution;
interaction;
model predictions;
flow where appropriate.

42. Use a table or figure only if it does a real cognitive job | Visual 不是 decoration

43. Do not show the same information three times | Text + table + figure 三重复制浪费 reader attention

Choose the best surface.

44. Prose should tell the reader what to notice | Text = highlighting layer

Manchester describes common Results writing as:

location/summary statement;
highlighting statement.

Example:

Table 2 summarises the primary outcomes. The largest between-group difference occurred in evidence integration, whereas sentence-level accuracy was similar.

45. Do not repeat every table cell | 逐格抄表是低价值 prose

Reader can already see 74.2, 68.1, 71.4, 69.8.

Tell them the pattern.

46. But do not force readers to extract every main result themselves | Table 不是“自己看”

State the main finding in text.

47. Figure captions should be interpretable | Caption 需要说明 group/measure/unit/uncertainty

48. Axes can distort | 图表比例也会产生 rhetoric

Truncated axes may exaggerate small differences.

49. Use log scales only when readers can interpret them | Scale choice 是 communication decision

50. Show uncertainty where relevant | Error bars / CI / credible intervals

51. Do not use decorative 3D charts | 视觉效果不能压过数据关系

52. Order categories meaningfully | Alphabetical 不一定是最佳 order

Order by:

natural sequence;
magnitude;
time;
pre-specified hierarchy.

53. Avoid colour as the only encoding | Accessibility / print / colour-vision differences

54. Report units everywhere they are needed | 分、秒、百分比、mg、odds ratio 都要明确

55. Significant figures should match measurement precision | 不要制造 fake precision

If measurement is to the nearest whole point, reporting 73.4287 may be misleading.

56. Percentages from tiny denominators can look too precise | 2/3 = 66.7%，但 n=3 更重要

57. Report counts with percentages where useful | n + % 常比 % alone 透明

58. Mean and median answer different distribution questions | 不要默认 mean 永远最好

Skewed data may need median/IQR.

59. Distribution can matter more than average | 两组同 mean，也可能 distribution 完全不同

60. Avoid hiding heterogeneity behind one average | Average effect 可能掩盖 subgroup or spread

61. But subgroup analysis has a high false-positive risk | Subgroup 不能随便挖

Distinguish:

pre-specified;
exploratory;
interaction tested;
descriptive only.

62. Do not claim subgroup difference because one group is significant and another is not | “A significant, B not significant” ≠ A differs from B

Test the interaction/difference directly where appropriate.

63. Correlations need direction and uncertainty | 不要只说 variables were related

Higher usage was moderately associated with higher score (r = …).

64. Correlation Results should not use causal verbs | Associated ≠ caused

65. Regression coefficients need interpretable units | coefficient 到底代表什么变化？

Example:

Each additional feedback cycle was associated with a 1.8-point higher revision score.

66. Odds ratios need reference clarity | OR 以谁为 reference？

67. Risk ratios need baseline context where practical meaning matters | Relative + absolute can both matter

68. Model fit statistics belong only if they help evaluate the model | 不要把所有 software output 搬进 Results

69. Report assumption failures or analytic consequences | Model problem 不能藏

70. Sensitivity analysis results show robustness | Sensitivity Results 说明结论对方法 choice 是否稳定

Example:

The estimate was similar after excluding two extreme observations.

71. Robustness is not an adjective without a test | “robust result” 要说经过什么 test

72. Multiplicity needs visible handling | 多 outcomes / tests 会改变 interpretation

Results should preserve which analyses were primary and which require caution.

73. Bayesian Results have different reporting language | 不要强套 p-value template

Depending on field:

posterior estimates;
credible intervals;
posterior probabilities;
Bayes factors.

74. Keep inferential framework internally consistent | 不要同一句混用 incompatible interpretations

75. Machine-learning Results need evaluation design clarity | ML Results 不只是 accuracy 最高

Report where relevant:

held-out test performance;
cross-validation;
baseline model;
class imbalance;
precision/recall/F1/AUC depending task;
uncertainty;
external validation.

76. Training performance is not test performance | 不要把训练集结果当 generalisation

77. Benchmark leakage can invalidate Results | Data leakage 是 Methods issue with Results consequence

78. Report baseline comparisons | New model needs meaningful comparator

79. Small benchmark gains may not be practically meaningful | Discussion evaluates meaning; Results reports magnitude

80. Qualitative Results have a different evidential surface | Qualitative Results 不是 numbers-lite

Primary unit may be:

theme;
pattern;
category;
case contrast;
process;
discourse feature.

81. Theme is not a topic label | Theme 要表达 analytic meaning

Weak:

Theme 1: Feedback.

Stronger:

Theme 1: Learners used feedback confidently only when the next action was explicit.

82. Qualitative Results need evidence for themes | Theme 后面要有 data trace

Use:

quotes;
fieldnote excerpts;
case descriptions;
document excerpts;

83. Quotes should illustrate, not replace analysis | Quote 不是 theme 自己

84. One vivid quote is not prevalence evidence | 生动 ≠ representative

85. Negative/deviant cases matter | 与 theme 不一致的 case 可以 refine theme

86. Frequency language requires methodological fit | most / many / several / few 要谨慎

In some qualitative traditions, frequency is secondary to meaning/structure.

87. Do not turn qualitative importance into pseudo-percentage | 不是所有 theme 都需要 63%

88. Show variation inside a theme | Theme 不应该把所有 participant 写成同一种人

89. Qualitative Results may include interpretation more tightly | 某些 qualitative genre 的 Results/Discussion 边界更融合

Still keep evidence and interpretive move traceable.

90. Mixed-methods Results require strand identity | Quant/Qual 不要混成一锅

Readers should see:

quantitative finding;
qualitative finding;
integration point.

91. Integration means relation, not averaging | Score + theme 不能假装同一单位

92. Joint displays can reveal mixed-methods relationships | Joint display 可以把两个 strand 对齐

Quant finding	Qual theme	Integrated inference
high performance gain	clear action prompts	action clarity may explain variation
low gain	feedback overload	dose may constrain benefit

93. Integrated inference usually needs Discussion-level caution | Joint display 可以在 Results，机制解释要 calibration

94. Systematic-review Results have their own structure | Review Results ≠ study Results

Common states:

study selection;
study characteristics;
risk of bias;
individual-study findings;
synthesis/meta-analysis;
heterogeneity;
certainty/limitations depending framework.

95. PRISMA flow is a Results object | How many records → included studies

96. Search yield is not the main result | 12,000 records found ≠ research conclusion

97. Meta-analysis needs effect + precision + heterogeneity | Pooled estimate alone may be insufficient

98. Forest plot is not self-explanatory | Text should highlight central synthesis

99. Heterogeneity is a result | 不要只放 I² 不解释其 reporting role

100. Publication-bias analyses have limits | Funnel plot 也不是 truth machine

101. Diagnostic-study Results need accuracy measures | Sensitivity, specificity, predictive values, likelihood ratios, AUC depending question

102. Predictive values depend on prevalence | 解释属于 Discussion，但 Results 需清楚 context

103. Time-to-event Results need censoring/at-risk clarity | Survival curve 需要 number-at-risk context

104. Economic-evaluation Results need cost + outcome relation | Cost-effectiveness 不是只报 cost

105. Repeated-measures Results need time × group logic | 不要只报每个 time point 独立 p-value

106. Interaction is itself a result | Difference-in-differences / interaction term may answer main question

107. Report unexpected findings | Unexpected 不代表不能写

But do not immediately explain them in Results.

108. Results can say the pattern was unexpected only if this status is relevant | Avoid emotionally loaded words

Elsevier advises avoiding subjective/emotional wording such as interestingly or unfortunately in Results.

109. “Interestingly” often hides why a result matters | 如果重要，直接说 pattern

Weak:

Interestingly, Group B improved more.

Better:

Group B showed the largest increase, particularly at delayed follow-up.

110. “Surprisingly” belongs only when expectation is explicit | Surprise 是 expectation relation，不是结果本身

111. Avoid evaluative adjectives | impressive / disappointing / excellent

112. Avoid causal adjectives | successful intervention

Results should report what happened.

113. Avoid mechanism verbs without design support | “because”, “therefore caused”, “resulted from” often cross into Discussion

114. But simple relational language is fine | “higher scores coincided with…” depending design

115. The Discussion boundary is functional, not lexical | 某个词不自动 illegal

The question is:

Am I describing what the evidence directly shows, or explaining why it happened and what it means beyond the immediate result?

116. Results can include limited highlighting | Highlighting = direct pattern visibility

The difference was concentrated in evidence integration.

This is still close to the result.

117. Results should avoid literature comparison unless genre combines sections | “consistent with Smith” 通常属于 Discussion

118. Results should avoid broad implication | “This has major implications for education” belongs later

119. Results should avoid policy recommendations | “Schools should adopt…” is not a Result

120. Results should avoid theoretical victory language | “This proves Theory A” is Discussion/Argument

121. Build a sentence-function audit | 每句话标记 job

Use tags:

LOC = locates table/figure/result;
OBS = direct observation;
COMP = direct comparison;
PREC = uncertainty/precision;
EXPL = explanation;
IMPL = implication.

EXPL/IMPL clusters in Results deserve inspection.

122. The “argument leakage” test | Argument leakage

Highlight words:

proves;
because;
therefore;
demonstrates that the mechanism;
should;
important;
consistent with previous research.

Some may be legitimate; inspect the move.

123. The “table duplication” test | Table duplication

Underline every number in the prose.

If the same number is already visible in a table and the sentence adds no pattern, cut it.

124. The “primary outcome visibility” test | Main answer 能否 15 秒内找到？

125. The “null honesty” test | 哪个结果最不支持 hypothesis？它还在吗？

126. The “secondary spin” test | positive secondary 是否盖过 primary null？

127. The “denominator” test | 每个百分比的 n 清楚吗？

128. The “precision” test | point estimate 有没有 uncertainty context？

129. The “comparison” test | compared with what?

130. The “time” test | immediate / delayed / follow-up time 是否清楚？

131. The “population” test | analysed population 是否与 enrolled population 相同？

132. The “missingness” test | missing outcome amount 是否可见？

133. The “exploratory” test | 事后 analysis 是否被读者误认为 pre-specified？

134. The “visual integrity” test | 图轴、scale、error bars 是否诚实？

135. The “qualitative evidence” test | 每个 theme 都有 supporting data trace 吗？

136. The “deviant case” test | 是否只选支持 theme 的 quotes？

137. The “mixed methods” test | quant 与 qual 是否保持 identity？

138. The “result lineage” test | 每个 result 都能回到 Method 吗？

139. Worked case: fictional structured-feedback trial | 完整案例：虚构 structured-feedback trial

All details below are fictional teaching material.

Question:

Does structured action-oriented feedback improve independent revision one week later among advanced bilingual learners?

Design:

96 learners, randomised 1:1.

Primary outcome:

unseen revision score at one week.

Secondary outcomes:

claim clarity, evidence integration, reasoning, organisation, sentence accuracy.

140. Weak Results version | 弱版本

The structured-feedback intervention was highly successful and clearly improved deep writing ability. Students in the intervention group performed significantly better, especially because the feedback helped them understand how to use evidence. Interestingly, the intervention was less effective for grammar, probably because these advanced students already had strong language skills. These results show that structured feedback should be adopted widely.

141. What is wrong with the weak version? | 问题

no actual estimate;
“highly successful” evaluative;
deep writing ability broader than measured outcome;
mechanism invented;
grammar explanation invented;
policy recommendation;
no null/uncertainty;
no denominators/precision.

142. Stronger primary-outcome paragraph | 示例

All 96 participants completed the immediate task, and 91 completed the one-week follow-up. On the primary unseen revision outcome, the structured-feedback group scored a mean of 74.2 (SD 8.6) compared with 68.1 (SD 9.1) in the standard-feedback group. The adjusted between-group difference was 6.0 points (95% CI 2.4–9.6).

143. What this paragraph does | 它只做 result job

follow-up denominator;
group means;
spread;
adjusted difference;
precision.

144. Stronger secondary-outcome paragraph | 示例

Table 2 shows the rubric-level outcomes. The largest between-group difference occurred in evidence integration (mean difference 1.4 points), followed by reasoning (0.9 points). Differences in claim clarity and organisation were smaller, and sentence-level accuracy was similar between groups.

145. What it does not say | 不解释 why

It does not say:

Evidence prompts caused deeper cognition.

That belongs in Discussion.

146. Stronger null-result paragraph | 示例

At 12-week follow-up, mean total scores remained higher in the structured-feedback group, but the estimated difference was smaller and imprecise (2.1 points, 95% CI −1.2 to 5.4).

147. This is not “no effect” | 它报告 evidence state，不夸大

148. Stronger exploratory paragraph | 示例

In a pre-labelled exploratory analysis, the one-week difference was larger among learners below the median baseline revision score than among those above it; the subgroup interaction was imprecisely estimated.

149. Good Results preserve hierarchy | Primary → secondary → delayed → exploratory

150. Discussion will later ask why | Results 先把 pattern 放在桌上

151. Practice A: repair p-value-only reporting | 练习 A

Repair:

The intervention significantly improved revision (p<.05).

152. Practice B: repair Results interpretation | 练习 B

Repair:

The higher score proves that explicit prompts deepen reasoning.

153. Practice C: preserve null result | 练习 C

Immediate effect positive; 12-week CI crosses zero.

Write one Results sentence.

154. Practice D: table highlighting | 练习 D

Table contains five outcomes.

Write one sentence telling the reader which pattern matters most.

155. Practice E: avoid table duplication | 练习 E

Prose currently repeats all 20 cells. Rewrite in one to two sentences.

156. Practice F: exploratory subgroup | 练习 F

Subgroup identified after seeing overall result.

Write transparent language.

157. Practice G: qualitative theme | 练习 G

Transform:

Theme: Feedback.

into an analytic theme.

158. Practice H: deviant case | 练习 H

Eight participants describe feedback as helpful; two describe it as overwhelming.

Write a theme sentence that preserves variation.

159. Practice I: mixed methods | 练习 I

Quant gain largest in classes with high fidelity; interviews describe clearer routines.

Write a result-level integrated statement without claiming mechanism as established.

160. Practice J: percentage denominator | 练习 J

Repair:

80% improved.

n=10.

161. Practice K: adverse trade-off | 练习 K

Accuracy improves, completion time increases 40%.

Write both results without evaluative language.

162. Practice L: visual integrity | 练习 L

Group means are 72 and 74 on a 0–100 scale. Axis starts at 71.

What risk does this create?

163. Model answers | 示范解析

A: Mean revision scores were 5.8 points higher in the intervention group than in the comparison group (95% CI 1.7–9.9).

B: The structured-feedback group achieved higher reasoning scores than the comparison group.

C: The intervention group scored higher immediately, but at 12 weeks the estimated between-group difference was smaller and imprecise.

D: The largest difference occurred in evidence integration, while sentence-level accuracy differed little between groups.

F: In an exploratory post-hoc analysis, the observed difference was larger among lower-baseline learners; this subgroup pattern was not pre-specified.

G: Learners used feedback confidently when comments identified a clear next revision action.

H: Most participants described action-oriented feedback as clarifying, although a minority experienced the volume of prompts as cognitively overwhelming.

I: Classes with higher implementation fidelity showed larger score gains, and interview data from the same classes more often described stable feedback routines.

J: Eight of 10 participants (80%) improved.

K: Accuracy increased by 6 percentage points, while mean completion time increased by 40%.

L: The truncated axis can visually exaggerate a small absolute difference.

164. The 20-minute Results drill | 20 分钟训练

Time	Task
3 min	map RQ → primary result
4 min	write magnitude + precision
4 min	write null/negative result
4 min	highlight table/figure pattern
5 min	argument-leakage + selective-reporting audit

165. The 45-minute growth session | 45 分钟增长模式

Time	Task
8 min	analyse target-journal Results sections
8 min	build RQ–Result matrix
10 min	draft primary/secondary results
10 min	table/figure highlighting
9 min	null/exploratory/spin audit

166. The 90-minute deep session | 90 分钟深度模式

Time	Task
15 min	target-venue + reporting guideline analysis
15 min	primary/secondary/exploratory hierarchy
20 min	full Results draft
15 min	tables/figures/captions
10 min	qualitative/mixed result treatment
15 min	selective-reporting + argument-leakage audit

167. Seven-day Results cycle | 七天训练循环

Day 1: primary outcome reporting.
Day 2: effect size + precision.
Day 3: null/negative findings.
Day 4: tables/figures.
Day 5: exploratory/subgroup analyses.
Day 6: qualitative/mixed-methods Results.
Day 7: full Results benchmark.

168. Twelve-week C1–C2 Results progression | 12 周路线

Weeks	Focus	Output
1–2	RQ alignment + outcome hierarchy	result maps
3–4	effect size, precision, nulls	quantitative result paragraphs
5–6	tables, figures, captions	visual-result packages
7–8	subgroups, sensitivity, exploratory analyses	transparent advanced Results
9–10	qualitative/mixed-methods reporting	theme/integration sections
11–12	discipline-specific independent transfer	submission-ready Results portfolio

169. Monthly benchmark | 每月基准任务

Choose one permitted or fictional study dataset and produce:

RQ–Result matrix;
primary-outcome paragraph;
secondary-outcome paragraph;
null/negative-result paragraph;
exploratory-analysis paragraph;
one table;
one figure;
two highlighting sentences;
selective-reporting audit;
table-duplication audit;
argument-leakage audit;
visual-integrity audit;
Method → Result lineage map.

170. First weak-link diagnosis | 第一个薄弱环节

Symptom	Likely weak link	Repair
Results = p-values	magnitude/precision missing	report estimate + interval
primary outcome hard to find	hierarchy weak	primary result first
all findings positive	selective reporting risk	audit null/adverse outcomes
prose repeats tables	highlighting weak	state pattern, not cells
Results explains mechanism	Discussion leakage	separate observation from explanation
subgroup looks definitive	exploratory status hidden	label + interaction/precision
qualitative themes are topic labels	analysis weak	write analytic theme + evidence
figure looks dramatic	visual scale distortion	audit axes/baseline/uncertainty

171. What not to do | 不要这样写 Results

Do not report only p-values.
Do not hide null, negative or adverse findings that change interpretation.
Do not let a positive secondary outcome replace a null primary outcome.
Do not repeat every table cell in prose.
Do not use visuals to exaggerate small differences.
Do not turn observational patterns into causal explanations.
Do not write policy recommendations inside Results.
Do not present post-hoc analyses as pre-specified.
Do not use one vivid qualitative quote as if it proves prevalence.
Do not confuse a clean story with complete reporting.

172. Research and reference floor | 研究与参考基础

University of Manchester Academic Phrasebank | Reporting Results — systematic reporting, tables/figures, highlighting statements, qualitative themes and the Results/Discussion boundary.
Elsevier | How to Write the Results Section — primary question first, clear objective reporting, negative results and non-duplicative tables/figures.
EQUATOR Network | Selecting the Appropriate Reporting Guideline — design-specific completeness requirements.
EQUATOR Network | Outcome Reporting Guidelines — outcome-focused reporting resources across study designs.
Purdue Writing Lab | Graduate Writers Guide — alignment between research methods, results and discussion.
Academic Phrasebank | Being Cautious — certainty calibration where Results statements approach interpretation.

173. Canonical eduKate routes | eduKate 主页面路由

174. 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, research results section, reporting research results, results vs discussion, effect size reporting, confidence intervals, null results, negative results, tables and figures research paper, statistical results writing, qualitative results writing, exploratory analysis reporting, subgroup analysis reporting, selective outcome reporting, C1 academic writing, C2 academic English, Results 怎么写, 研究结果怎么写, null result 怎么写, p value 怎么写, effect size 怎么写, table figure 学术英语, 中文母语学术英语 and C1 C2 research writing.

175. The one-page Results operating system | 一页 Results 操作系统

Map every research question to a result.
Report the primary outcome first.
Preserve primary/secondary/exploratory hierarchy.
Report direction, magnitude and precision.
Do not use p-values as the result.
Report null, negative and adverse findings honestly.
Keep denominators and analysed population visible.
Use tables for structured comparison.
Use figures for patterns/trends/distributions.
Use prose to highlight, not duplicate.
Label subgroup/exploratory/post-hoc analyses.
Preserve uncertainty and missingness.
For qualitative Results, write analytic themes + evidence.
For mixed methods, preserve strand identity before integration.
Stop before mechanism, literature comparison and policy implication dominate.
Run null-honesty, secondary-spin, table-duplication and visual-integrity audits.

176. Final assignment | 最终作业

Choose a permitted or fictional study dataset.

Complete this chain:

Write the research questions.
Identify primary, secondary and exploratory outcomes.
Build an RQ–Result matrix.
Write the primary-outcome paragraph.
Include direction, magnitude and precision.
Write the most important null/negative finding.
Write one adverse/trade-off result if present.
Create one table.
Create one figure.
Write one highlighting sentence for each.
Write one exploratory-analysis paragraph with transparent labelling.
If qualitative, write one analytic theme with supporting evidence and one deviant case.
If mixed methods, create a joint display.
Run the p-value-only audit.
Run the primary-outcome visibility audit.
Run the secondary-spin audit.
Run the table-duplication audit.
Run the argument-leakage audit.
Run the visual-integrity audit.
Map every major Result back to its Method parent.

Then ask:

If a reader disagreed with my eventual Discussion, would they still agree that my Results section represented the observed evidence fairly?

如果一个 reader 不同意我后面的 Discussion，他仍然会不会认为我的 Results 公平、完整地呈现了 observed evidence？

If yes, the Results section is doing its job.

If no, the argument may already be shaping the evidence before the Discussion begins.

Continue | 继续

The next lesson will take the findings into the Discussion: how to interpret results without repeating them, inventing mechanisms or turning limitations into ritual apologies.

下一课进入 Discussion 深层写作：怎样解释结果而不重复 Results、不编造 mechanism，也不把 limitations 写成形式化道歉。

Next: EDKS-ADV-ZH-0020 · Lesson No.020 · Write a Discussion That Explains Without Overclaiming · 把讨论写成解释，而不是把结果说得更大

Back to Advanced English Chinese Edition Hub · 返回高级英语中文版主页

Small Group Tutorials

How to Learn Advanced English (Chinese Edition) | Lesson No.019 | Write Results That Reveal the Pattern Without Arguing Ahead of the Evidence | 第019课：把研究结果写清楚，而不是提前替证据下结论

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

Before we begin: Lesson 015 owns section architecture; Lesson 019 owns Results depth | 第015课管 section architecture；第019课深入 Results

1. Manchester: Results should present findings systematically and in detail | Results 首先要 systematic + detailed

2. Elsevier: primary question first, then secondary findings | 先回答最重要的问题

3. EQUATOR: reporting completeness depends on study design | Results 也有 design-specific reporting floor

4. Data are not Results | Data ≠ Results

5. Results already contain analytical selection | Results 不是“零 interpretation”

6. The Results section should answer the research question visibly | RQ → Result 必须直接

7. Build an RQ–Result matrix before drafting | 先建立 RQ–Result matrix

8. If a research question has no result, something is wrong | Question 没有 answer = architecture problem

9. If a major result has no research-question parent, inspect it | Orphan Result

10. Primary outcome first | Primary outcome 先出现

11. Primary outcome is not “the result that worked best” | Primary 不能事后挑

12. Secondary outcomes enrich; they do not replace the primary story | Secondary 负责补充

13. Exploratory analyses must be labelled | Exploratory 不应伪装 confirmatory

14. Exploratory does not mean useless | Exploratory 可以产生下一步 hypothesis

15. Report direction, magnitude and precision | 不要只报 “significant”

16. “Significant” without an estimate hides the result | 只有 p-value，不够

17. Statistical significance is not effect magnitude | p-value 不能替代 effect size

18. Practical importance is not a Results adjective | “important effect” 通常属于 Discussion

19. Confidence intervals are high-information language | CI 可以同时显示 direction + magnitude + precision

20. Wide intervals deserve visible wording | Precision 不足要可见

21. Null results are results | Null result 不是空白

22. Non-significance is not equivalence | 不显著 ≠ 相同

23. “No evidence of X” is not always “evidence of no X” | absence of evidence ≠ evidence of absence

24. Report negative findings honestly | Negative result 不能因为不好看就删

25. Selective outcome reporting creates a distorted study | Selective reporting 改变 study identity

26. Preserve pre-specified hierarchy | Outcome switching 是高风险

27. Adverse findings may be more important than positive outcomes | Harm/trade-off 不能因为篇幅删掉

28. Denominators matter | 百分比没有 denominator 很容易误导

29. Flow counts matter in longitudinal/trial work | 谁进入、谁完成、谁被分析

30. Attrition should be visible | Attrition 不是 footnote

31. Missingness belongs in the result state when it affects interpretation | Missing data amount 是 result-adjacent information

32. Baseline description has a job, not a ritual | Baseline table 不是装饰

33. Organise Results by research logic | Results 顺序不一定按 analysis 软件输出

34. Software output order is not reader order | SPSS/R 输出顺序不是文章结构

35. Use subheadings when they map the evidence | Subheading 要帮助 reader 看问题

36. Avoid result subheadings that already interpret | 不要标题先下结论

37. Begin each subsection with the question or finding state | 先定位 reader

38. Then report the result directly | 不要绕三句才说数字

39. Highlight, do not narrate every calculation | Results 不是 analysis diary

40. Tables are for structured comparison | Table 擅长让读者查数

41. Figures are for visible patterns | Figure 擅长趋势、关系、分布

42. Use a table or figure only if it does a real cognitive job | Visual 不是 decoration

43. Do not show the same information three times | Text + table + figure 三重复制浪费 reader attention

44. Prose should tell the reader what to notice | Text = highlighting layer

45. Do not repeat every table cell | 逐格抄表是低价值 prose

46. But do not force readers to extract every main result themselves | Table 不是“自己看”

47. Figure captions should be interpretable | Caption 需要说明 group/measure/unit/uncertainty

48. Axes can distort | 图表比例也会产生 rhetoric

49. Use log scales only when readers can interpret them | Scale choice 是 communication decision

50. Show uncertainty where relevant | Error bars / CI / credible intervals

51. Do not use decorative 3D charts | 视觉效果不能压过数据关系

52. Order categories meaningfully | Alphabetical 不一定是最佳 order

53. Avoid colour as the only encoding | Accessibility / print / colour-vision differences

54. Report units everywhere they are needed | 分、秒、百分比、mg、odds ratio 都要明确

55. Significant figures should match measurement precision | 不要制造 fake precision

56. Percentages from tiny denominators can look too precise | 2/3 = 66.7%，但 n=3 更重要

57. Report counts with percentages where useful | n + % 常比 % alone 透明

58. Mean and median answer different distribution questions | 不要默认 mean 永远最好

59. Distribution can matter more than average | 两组同 mean，也可能 distribution 完全不同

60. Avoid hiding heterogeneity behind one average | Average effect 可能掩盖 subgroup or spread

61. But subgroup analysis has a high false-positive risk | Subgroup 不能随便挖

62. Do not claim subgroup difference because one group is significant and another is not | “A significant, B not significant” ≠ A differs from B

63. Correlations need direction and uncertainty | 不要只说 variables were related

64. Correlation Results should not use causal verbs | Associated ≠ caused

65. Regression coefficients need interpretable units | coefficient 到底代表什么变化？

66. Odds ratios need reference clarity | OR 以谁为 reference？

67. Risk ratios need baseline context where practical meaning matters | Relative + absolute can both matter

68. Model fit statistics belong only if they help evaluate the model | 不要把所有 software output 搬进 Results

69. Report assumption failures or analytic consequences | Model problem 不能藏

70. Sensitivity analysis results show robustness | Sensitivity Results 说明结论对方法 choice 是否稳定

71. Robustness is not an adjective without a test | “robust result” 要说经过什么 test

72. Multiplicity needs visible handling | 多 outcomes / tests 会改变 interpretation

73. Bayesian Results have different reporting language | 不要强套 p-value template

74. Keep inferential framework internally consistent | 不要同一句混用 incompatible interpretations

75. Machine-learning Results need evaluation design clarity | ML Results 不只是 accuracy 最高

76. Training performance is not test performance | 不要把训练集结果当 generalisation