Prompt Details
Model
Chat - (gpt-5.6)
Token size
1,630
Example input
[PUBLICATION / PLAIN-LANGUAGE]: PUBLICATION
[e.g. "randomized controlled trial", "observational cohort study", "cross-sectional survey", or "not specified"]: randomized controlled trial
[e.g. clinical medicine, psychology, ecology, or "not specified"]: clinical medicine
[PASTE RESULTS HERE]: Independent samples t-test comparing systolic blood pressure between treatment group (n=45, M=128.4, SD=9.2) and control group (n=43, M=134.1, SD=8.7). t(86)=2.94, p=0.004. Cohen's d=0.63, 95% CI (1.8, 9.6) mmHg difference.
Example output
PUBLICATION style, clean significant result (RCT)
Input results:
STUDY DESIGN: randomized controlled trial Independent samples t-test comparing systolic blood pressure between treatment group (n=45, M=128.4, SD=9.2) and control group (n=43, M=134.1, SD=8.7). t(86)=2.94, p=0.004. Cohen's d=0.63, 95% CI [1.8, 9.6] mmHg difference.
Output:
Narrative write-up: Participants receiving the treatment showed significantly lower systolic blood pressure (M=128.4, SD=9.2, n=45) compared to controls (M=134.1, SD=8.7, n=43), t(86)=2.94, p=0.004, d=0.63, 95% CI [1.8, 9.6] mmHg. This represents a medium-to-large effect size.
Wording accuracy check: Since this is a randomized controlled trial, causal language (e.g. "the treatment reduced blood pressure") is appropriate here, unlike in observational designs. Effect size and CI were reported alongside the p-value, which is correct practice — no overclaiming detected.
What's missing: Consider reporting whether this was a primary or secondary outcome, and whether multiple-comparisons correction was applied if other outcomes were also tested.
By purchasing this prompt, you agree to our terms of service
GPT-5.6
Turn raw statistical output (R/SPSS/Python results, p-values, effect sizes, model coefficients, odds/hazard ratios) into properly worded narrative text — either publication-style for a paper's Results section, or plain-language for a non-technical report. Flags overclaiming, missing effect sizes, causal language where only correlation was tested, and common interpretation traps (percentage-points vs. percent, standardized vs. raw coefficients, log-scale outcomes, HR/OR misread as probability).
...more
Updated 3 weeks ago
