RESEARCH PAPER SECTIONS

Find out which statistics a paper used and what each result means

The statistical analysis subsection tells you which tests and models produced the numbers in the results: the comparison used, the variables adjusted for, how missing data was treated and how multiple outcomes were handled. It is often written in compressed technical language. Ask Search+ "Which statistical methods were used for the main outcome, and what did the model adjust for?" and the answer explains it in plain terms with a citation, so you can read the authors' exact description alongside.

Start a workspace Read about research papers

Creating an account needs no payment.

Last updated October 2026

How to read a paper's statistics with Search+

Upload the paper and its statistical supplement
Detailed model specifications and sensitivity analyses often sit in supplementary material. Add it to the workspace so questions can reach it.
Start from the main outcome
Ask which analysis produced the primary result, then ask what it adjusted for and what assumptions it relied on. Secondary analyses can wait.
Ask what each reported value means here
Ask what a specific coefficient, ratio or interval means for this study. The answer explains it from the paper and cites the passage.

Questions to ask about statistical methods

Main test

Which statistical test or model was used to analyse the primary outcome?

Adjustment

Which variables were included as covariates or confounders, and why?

Missing data

How was missing data handled: complete-case analysis, imputation or another method?

Multiple comparisons

Did the authors test many outcomes or subgroups, and did they adjust for multiple comparisons?

Sensitivity analyses

Were sensitivity analyses run, and did they change the main result?

Interpreting a value

What does the reported hazard ratio mean for the comparison in this study?

What to look for in a statistical analysis section

The test should fit the data

The choice of test depends on the type of outcome and the design, for example continuous versus binary outcomes or paired versus independent groups. The paper should say which it used for each outcome.

P-values need context

A p-value describes how surprising the data would be if there were no real effect. It is not the probability that the finding is true, and it says nothing about the effect's size, so read it beside the estimate and its interval.

Adjustment choices change results

Which confounders a model includes can move an estimate a lot. Look for how the authors chose them and whether results are shown with and without adjustment.

Many tests, more false positives

Testing many outcomes or subgroups raises the chance of a result that looks significant by chance alone. Check whether the authors correct for this or label such results exploratory.

Explanations tied to the paper

Search+ answers from the analysis section with inline citations to the supporting excerpt. It reports what the authors did and does not re-run or audit the analysis.

Statistics wording you may see, and what to ask next

Wording you may see	What it usually signals	A follow-up question
"adjusted for age, sex and"	A multivariable model with covariates	Is an unadjusted result also reported, and how different is it?
"95% confidence interval"	The precision of an estimate	How wide is the interval for the main result?
"multiple imputation"	Missing values estimated from other data	How much data was missing before imputation?
"Bonferroni" or "false discovery rate"	A correction for multiple comparisons	Which results remain after correction?
"mixed-effects model"	A model for repeated or clustered data	What was treated as the clustering unit?
"sensitivity analysis"	A check on whether results hold under other assumptions	Did any sensitivity analysis change the conclusion?

What are the statistical methods of a research paper?

The statistical methods are the part of a paper, usually a subsection of the methods, that describes how the data was analysed: which tests or models were used, what was adjusted for, how missing data and multiple comparisons were handled, and what threshold was used to judge results.

Statistical methods are not the results. The methods say how the analysis was done; the results report what it produced. Reading them together tells you whether a reported number came from the planned analysis.

Statistical methods questions

Can Search+ explain the statistics used in a paper?

Yes. Ask which tests or models were used and what a reported value means, and Search+ explains it from the paper with citations to the analysis section.

Will it check whether the right test was used?

No. It reports and explains what the authors did, with citations. Deciding whether the analysis suits the data needs your own statistical judgment or a statistician's.

Can it explain a value such as an odds ratio in plain language?

Ask what the value means for the groups compared in this study. The answer explains it in context and cites the result it refers to.

Can I compare analysis methods across studies?

Yes. Put the papers in one workspace and ask which methods and adjustments each one used. The answer cites each paper, which can explain why their estimates differ.

Does it read equations in the paper?

It can answer from equations in the paper, but symbols, subscripts and Greek letters are easy to garble, so read the cited passage in the original before you reproduce a model.

Can I rely on its statistical explanations?

Use them to orient yourself, and check the cited text. AI answers can be incomplete or wrong, especially on technical detail.

Related parts of a research paper

Research papers

Read one paper with cited answers

Sample size

Power and participant numbers

Key findings

What the analysis produced

Methodology

The design behind the analysis

Datasets