How to write up quantitative results in a dissertation

Q: How do I report a p value in APA 7?

In italics, with no leading zero, exact to two or three decimals: p = .004. A p value below .001 is written p < .001, never p = .000.

Q: Should the text repeat what a table shows?

No. APA's handout says not to repeat statistics in both. Name the main result and point to the table.

Q: Do I need effect sizes and confidence intervals in a dissertation?

APA's reporting standards ask for both for each test where possible, and Capella requires effect size and power. Include them.

Q: Where do I explain what the results mean?

In the discussion. Capella, Westminster and USC keep interpretation out of the results, though USC allows a direct comparison with an earlier study.

Q: Can EdCitation check my results chapter?

It checks the references, free: EdCitation's Verify references marks each entry verified, doubtful or not found. On Pro, Mechanics QA flags numbers that disagree between sections. It never runs, checks or writes your statistics.

Write a quantitative results chapter: its order, statistics in APA 7, p values, effect sizes, tables and figures, missing data, CONSORT and STROBE, with an example.

The EdCitation editorial teamWritten and checked against the style publishers’ own guidance and the studies cited. Last checked September 26, 2026. How we check

A quantitative results chapter shows, in a fixed and checkable order, what the analysis found for each question the dissertation asked. Writing up quantitative results well is mostly discipline about order, numbers and restraint. EdCitation publishes this as the quantitative results guide of How to write a dissertation; the qualitative chapter has its own guide.

Every rule below was read at its source on 26 September 2026: APA Style's pages on statistics, tables and figures, APA's reporting standards, the CONSORT and STROBE sites, and the dissertation guides of Capella, Southampton, the University of Southern California (USC) and Westminster. Where they disagree, we say so.

EdCitation never writes any part of a dissertation and never analyses data. It finds the papers behind your measures with Find sources, builds their references with Cite a source, and checks the chapter's list with Verify references, all free.

What goes in a quantitative results chapter, and in what order?

The sample, the descriptive statistics, each hypothesis in the order posed, anything unplanned, then a summary. Capella University (2024) sets four sections: a description of the sample, the hypothesis testing, a summary of that testing, and any post hoc analyses. University of Westminster Library (2025) asks for descriptive statistics such as means and standard deviations before the inferential statistics.

In steps:

Restate the questions or hypotheses and say how the chapter is laid out; Capella wants nothing repeated from the method.
Describe the sample analysed: how many were invited, responded, withdrew or were excluded, and why.
Report how much data was missing and what you did about it.
Give descriptive statistics for every variable, overall and by group.
Report each test's assumption checks and what you did when one failed.
Take each hypothesis in turn: the test, the result in APA form, the effect size and interval, and whether it was supported.
Report unplanned analyses under their own heading.
Sum up the answer to each question, without explaining it.

Where the guides allow a different order

University of Southampton Library (2026) lists four structures: chronological, as for a pre-test and post-test design; by method, in mixed methods; by question or hypothesis; or by variable. USC Libraries (2026) offers a synopsis then the key findings, or each result in turn then a synopsis, and asks you to use one. Your handbook's order, if it gives one, wins.

How do I report statistics in APA 7 style?

Symbol in italics, degrees of freedom in parentheses, the value, the exact p value and an effect size, with spaces around the equals sign. American Psychological Association (2024a) sets out the rules in its numbers and statistics handout, and APA's sample tables attach the degrees of freedom to the symbol, as in t(40) and F(1, 294) (American Psychological Association, 2024c).

Element	APA 7 rule	Right	Wrong
Statistical symbols	Latin letters in italics, Greek letters upright	M, SD, t, F, p, R², α, η²	M, SD, α
Operators	A space either side; a minus sign takes a space before only	M = 7.7; –8.25	M=7.7
Leading zero	Only when the statistic can exceed 1	d = 0.41; r = .32; p = .004	r = 0.32
Decimals, integer scales	Means and SDs of questionnaire scores to one decimal	M = 9.4	M = 9.4318
Decimals, other statistics	Correlations, proportions, t, F and chi-square to two decimals	t(210) = 2.92	t = 2.917235
p values	Exact, to two or three decimals; below .001 as p < .001	p = .03; p < .001	p = .000; p < .05
Words or symbols	The word in a sentence, the symbol beside an operator	"the means were"	"the Ms were"

Symbols such as M, SD, t, df, p and N need no definition; abbreviations such as ANOVA and CI do. The decimal rules are general: round as far as later use and precision allow.

Exact p values, effect sizes and confidence intervals

APA's reporting standards ask for every inferential test run, the exact p value, enough statistics, such as the degrees of freedom, to reconstruct each test, and an effect size with a confidence interval where possible (American Psychological Association, 2024b; Appelbaum et al., 2018). Capella adds the alpha level, the power obtained and whether each null hypothesis was accepted or rejected.

For intervals in a table, APA's sample regression table gives the limits their own columns, LL and UL, and allows square brackets in one column instead.

Primary, secondary and exploratory results

The standards keep primary, secondary and exploratory hypotheses visibly apart. For an unplanned analysis, Capella asks what prompted it, what you ran and what it showed; that is not the same as the post hoc tests that follow an ANOVA inside planned testing.

How do I report missing data and assumption checks?

Say how much is missing, why, and what you did. APA's standards ask for the share of missing data, the evidence for its cause, such as missing completely at random, missing at random or missing not at random, and the method used, such as listwise deletion or multiple imputation.

STROBE's item 14 asks for the number missing on each variable (STROBE Initiative, n.d.); CONSORT 2025's item 26 asks, for each outcome and group, how many were analysed and why data are missing (Hopewell et al., 2025).

Assumptions, and what you did when one failed

Capella asks for each test's assumptions, how you checked them, the result and what you did if they failed; a transformed variable gets descriptive statistics both raw and transformed. Report a Levene's test for homogeneity of variance, or a check of the normality assumption, with its result. If a failed check sent you to Welch's t-test or a non-parametric test, say so there.

Should I use a table or a figure, and what goes in the text?

A table for exact values, a figure when the pattern matters more, and in the text a sentence saying what each shows. American Psychological Association (2021b) defines a table as rows and columns; every other display, from bar graph to scatterplot, is a figure (American Psychological Association, 2021a). Each should make sense without the text.

The handout says not to repeat statistics in both text and table. Southampton warns against the opposite: a bare "look at Table 1" is not an account of the data. So name the main result and point to the table for the rest.

Setting up the table or figure

Number each in order of first mention, the number in bold, an italic title-case title below it. Use horizontal lines only, with abbreviations and asterisks explained in notes beneath.

The handout asks for exact p values in tables too, with <.001 for the smallest; APA's sample correlation and ANOVA tables show the alternative, asterisks explained in a probability note (American Psychological Association, 2024c). Use one method per table.

Where the university and APA differ

Southampton says figure captions normally go below the figure, and to confirm with your supervisor; APA puts a figure's number and title above it. Under APA 7, APA wins; under a departmental template, the template does.

What does not belong in the results chapter?

Explanation, speculation and the literature, in most guides' view. Westminster says describe objectively and leave interpreting to the discussion; Capella keeps even the sample's limits for chapter five. The comparison with earlier studies is in the discussion chapter guide, and the choice of tests belongs in the methodology chapter.

USC Libraries (2026) differs: it allows a result to be compared with a specific earlier study. Follow your handbook. Three more things stay out:

Raw data. USC, Southampton and Westminster send it to an appendix, if anywhere.
Words that judge. USC warns against vague phrases such as "appeared to be greater", and notes that results prove nothing; they support or challenge a hypothesis.
Silence about bad news. USC and Southampton say report negative and unexpected results; Westminster adds that misrepresenting data has ethical implications.

USC also asks for the past tense.

Which reporting guideline applies: CONSORT, STROBE or APA's JARS?

CONSORT for a randomised controlled trial, STROBE for a cohort, case-control or cross-sectional study, and APA's JARS–Quant in psychology. A dissertation is not a journal article, but each checklist lists what a reader needs; ask whether your department expects one.

Guideline	Covers	What it asks of the results
CONSORT 2025, 30 items and a flow diagram	Randomised trials	Flow diagram; recruitment dates; baseline table by group; numbers analysed; effect sizes with precision, absolute and relative for binary outcomes; harms; pre-specified or post hoc analyses
STROBE, 22 items	Cohort, case-control and cross-sectional studies	Numbers at each stage, with reasons; characteristics and missing data per variable; unadjusted and adjusted estimates with 95% confidence intervals; subgroup and sensitivity analyses
JARS–Quant	Quantitative research in psychology	Participant flow; missing data; descriptives by group; every test with exact p values, effect sizes and intervals

CONSORT 2025 is the current version (Hopewell et al., 2025). STROBE's authors limited it to three observational designs (von Elm et al., 2007), and its site says the checklist is not an instrument for judging a study's quality.

A worked example: one hypothesis reported badly, then well

The example is ours, and its data are invented: an imagined cross-sectional survey testing whether undergraduates in paid work for 15 or more hours a week score higher for anxiety on the GAD-7.

The weak version

Table 2 shows the results. Students who work had a mean anxiety score of 9.4318 and non-workers 7.5968. An independent samples t-test was highly significant (t=2.92, P=0.004) proving that part-time work causes anxiety. This is probably because working students have less time to study.

It repeats the table, gives a questionnaire mean to four decimals, omits standard deviations, degrees of freedom, effect size and interval, sets symbols upright without spaces, gives p a capital and a leading zero, claims cause from a survey, explains the result, and says nothing about who was left out.

The better version

Of 220 students who returned the survey, 8 (3.6%) left one or more GAD-7 items blank and were excluded, leaving 212 for analysis. Internal consistency of the GAD-7 in this sample was α = .89. Students working 15 or more hours a week (n = 88) reported higher anxiety (M = 9.4, SD = 4.6) than those working fewer hours or none (n = 124; M = 7.6, SD = 4.3). Levene's test showed no difference in variances, F(1, 210) = 0.86, p = .355, so an independent-samples t-test was used. The difference was statistically significant, t(210) = 2.92, p = .004, with a mean difference of 1.8 points, 95% CI [0.58, 3.02], and an effect size of d = 0.41. Hypothesis 1 was supported.

Every number needed to check the test is there: α has no leading zero because it cannot exceed 1, while Cohen's d keeps one because it can. Why working students might be more anxious waits for the discussion.

How EdCitation helps with a quantitative results chapter

EdCitation is the best tool for a results chapter's references, because every entry is looked up in the publisher's record rather than composed and a scale's validation study or a statistics text is easily copied, wrongly, from another paper's list. We ran each free tool on 26 September 2026.

Find sources searches about 300 million published works. "GAD-7 validation university students" returned 29,122 results, led by GAD-7 validation studies among students in Bangladesh (PLoS ONE, 2021), Portugal (Cadernos de Saúde Pública, 2017) and Spain (PeerJ, 2022). The fourth was a prevalence survey, not a validation study: read before citing.

Cite a source builds the entry from a DOI, title, PubMed ID or ISBN in six styles, screened for retraction. From DOI 10.1001/archinte.166.10.1092 it returned on 27 September 2026: Spitzer, R. L., Kroenke, K., Williams, J. B. W., & Löwe, B. (2006). A brief measure for assessing generalized anxiety disorder: The GAD-7. Archives of Internal Medicine, 166(10), 1092–1097. https://doi.org/10.1001/archinte.166.10.1092. The publisher's record gives only the first page; the range came from PubMed. The title is in sentence case, as APA 7 asks. From ISBN 9781526419521 it returned "Field, A. (2017). Discovering statistics using IBM SPSS statistics. SAGE Publications." with no edition: add it from your copy's title page, and restore the capital in the software's name, IBM SPSS Statistics.

Verify references checks a pasted list or uploaded paper. We gave it the corrected GAD-7 entry, the Appelbaum et al. (2018) standards dated 2016, and a paper we invented. The verdicts, exactly: the GAD-7 paper verified ("The DOI resolves to this record and the title matches."); the reporting standards doubtful ("The DOI and title are right, but the work was published in 2018, not 2016."); and the invented paper not found ("No publisher's or registry's record matches this reference."). One it cannot reach shows "could not check", never not found.

References from a file checks every reference in the uploaded chapter and matches every citation to the list. Mechanics QA checks the file against the APA 7 student paper's formatting and finds figures that do not add up or disagree between sections, such as 212 participants in one place and 214 in another. Each gives three checks a month with a free account; Pro, $8 a month, lifts the limit. Check your paper reads your handbook into a checklist, free; Theoretics QA (Max, $24 a month) judges the draft against its instructions and rewrites nothing. See Pricing.

None of it runs or checks your analysis: Mechanics QA can see that two numbers disagree, not which is right.

Quick questions

How do I report a p value in APA 7?