What Should You Do When You Don’t Understand the Methods or Analysis?

What Do You Do When the Paper Suddenly Stops Making Sense?

You are following a research paper reasonably well until you reach the methods or analysis. Then the terminology changes. The authors mention a sampling procedure you have never encountered, a statistical model you cannot interpret, a qualitative analytical approach you do not know, or a technical procedure described as though every reader should already understand it.

It is tempting to skip ahead to the results and trust that the authors used the method correctly. The opposite reaction is equally common: deciding that you cannot understand the paper at all because one technical section is beyond your current expertise.

Neither response is particularly useful.

You do not need specialist-level mastery of every technique used in every paper you read. You do need to recognize when something you do not understand is important enough that you cannot responsibly interpret the paper without resolving it.

Identify the Exact Gap, Then Decide Whether It Matters

You Do Not Need to Master the Method, but You Cannot Treat It as a Black Box

First, identify what you actually do not understand

"I don't understand the methods" is often too broad to be useful.

You may understand the study design but not the statistical analysis. You may understand what the analysis does but not why a particular variable was transformed. Perhaps the method itself is familiar, but one parameter, assumption, coding procedure, sampling decision, or model output is not.

Turn the vague difficulty into a specific question.

A specific question is much easier to investigate than an entire methodology.

Determine whether the gap is consequential

Not every unresolved technical detail deserves the same effort.

Suppose you encounter an unfamiliar algorithm in a paper you are reading only to understand the general research problem. A broad explanation of what the algorithm does may be enough. If your own conclusion depends on an estimate produced by that algorithm, however, the threshold changes.

The second type requires action.

This distinction follows the broader principle that the depth of understanding you need should depend on how you intend to use the paper. A method can remain partly unresolved during initial screening and become essential later when the paper moves into your argument or methodology.

Return to the research question before studying the technique

Technical methods make more sense when you know what problem they are supposed to solve.

Before searching for a tutorial on an unfamiliar analysis, ask what the researchers wanted to learn. Were they comparing groups, estimating change over time, examining an association, identifying latent dimensions, classifying observations, interpreting experiences, modeling nested data, synthesizing previous studies, or estimating some other quantity?

Then ask why the method was used for that purpose.

Carey, Steiner, and Petri recommend asking what the authors wanted to know, what they did, and why they did it that way. This sequence is particularly useful when a technical method is unfamiliar because it reconnects the technique to the research problem rather than leaving it as a collection of terminology.

You may discover that you understand more of the analysis than its technical name initially suggested.

Find the minimum methodological understanding you need

When a method is consequential, you usually need answers to several practical questions before you need mathematical or procedural mastery.

Ask:

• What is the method designed to do?
• Why is it appropriate, or potentially appropriate, for this research question and data?
• What information goes into it?
• What output does it produce?
• How should the relevant output be interpreted?
• Which assumptions or methodological choices could materially affect that interpretation?
• What does the method not allow the researchers to conclude?

For many reading purposes, answering those questions gives you considerably more value than attempting to reproduce every calculation or procedural step.

If you intend to use the method yourself, replicate the study, conduct peer review, or make a detailed methodological critique, the required depth is substantially greater.

Read around the difficult sentence before leaving the paper

Before opening another source, inspect the paper itself more carefully.

Authors may define the method when it first appears, explain why it was chosen, describe its implementation elsewhere in the methods, or clarify the relevant output in a table note, figure caption, appendix, or supplementary file.

Look backward and forward. Which variables or materials are involved? Is the unfamiliar procedure attached to one particular research question? Does a later paragraph explain what the resulting estimate means?

Scientific papers are dense enough that an explanation can easily be separated from the sentence that triggered your confusion.

Also inspect tables and figures. Carey and colleagues recommend unpacking figures and tables and repeatedly returning to the methods to understand how the presented data were obtained. The reverse can also help: seeing the output may clarify what the method was intended to produce.

Check supplementary materials before assuming the explanation is missing

Journal word limits and disciplinary conventions can push important methodological detail outside the main article.

Supplementary files may contain questionnaires, coding frameworks, robustness checks, analytical specifications, extended equations, preprocessing procedures, intervention materials, additional tables, sensitivity analyses, or other information necessary for understanding what was done.

Carey and colleagues explicitly recommend consulting supplementary material when necessary to understand a scientific work. StatPearls similarly treats the methods as the primary reference for replication and recommends returning to them when critical appraisal raises questions about study design or analysis.

If the paper matters to your work, do not stop at the main PDF merely because the technical detail you need is not immediately visible there.

Follow the methodological citation

Researchers often describe established procedures briefly and cite the source where the method was introduced or explained in detail.

You might see statements such as "analysis followed the procedure described by..." or "scores were calculated according to..." followed by a citation. If that method is central to your interpretation, the cited source may be exactly where you need to go next.

Prioritize citations attached directly to the methodological step you do not understand.

You do not necessarily need to read the entire cited paper. Find the definition, procedure, assumptions, or interpretation necessary to answer your question, then return to the focal study.

This is usually more efficient than trying to learn an entire methodological field because one term was unfamiliar.

Use an authoritative methods source when the cited paper assumes too much

The original methodological paper is not always the easiest place to learn a technique. It may be written for specialists and assume substantial prior knowledge.

An authoritative textbook, methodological handbook, scholarly tutorial, professional society resource, or well-established educational resource may provide a clearer entry point.

The key is to choose sources that explain the method accurately rather than relying on whatever simplified explanation happens to rank first in a search engine.

For statistical techniques, documentation from reputable statistical organizations, textbooks, or peer-reviewed methodological literature can help. For reporting and study-design questions, resources collected by the EQUATOR Network may point you toward guidelines specific to particular research designs.

Once you understand the basic method, return to the actual paper. Generic explanations cannot tell you whether the authors implemented the method appropriately in this particular study.

If statistics are the problem, translate the output before learning the mathematics

An unfamiliar statistical method often looks harder because the notation arrives before the meaning.

Start with the output you need to interpret.

What is the outcome? Which predictor, group, or comparison matters? Is the reported quantity a mean difference, correlation, regression coefficient, odds ratio, hazard ratio, standardized effect, model fit statistic, probability, or something else?

Then ask what a larger, smaller, positive, negative, zero, or null value would mean.

You may not need to derive the estimator to understand the substantive result.

For example, understanding a regression finding may initially require knowing which variable is the outcome, what a particular coefficient represents, what other variables were included in the model, and how uncertainty around the coefficient is reported. That is different from knowing how to derive the regression estimator from matrix algebra.

If statistical notation is the main obstacle, use a focused approach to reading the statistical result in terms of its comparison, estimate, magnitude, and uncertainty.

Do not reduce an unfamiliar analysis to whether p is below.05

When readers do not understand an analysis, the p-value can become an attractive escape route. It appears to offer a simple verdict: significant or not significant.

That is not enough.

The American Statistical Association has cautioned against using p-values as a substitute for scientific reasoning. A p-value does not tell you the size or practical importance of an effect, whether the model was appropriate, whether important assumptions were met, or whether the study design supports the interpretation.

If you cannot explain what was estimated or tested, knowing only that p <.05 does not solve the comprehension problem.

For qualitative analysis, ask how the researchers moved from data to interpretation

Not all difficult analyses are statistical.

A qualitative paper may refer to thematic analysis, grounded theory, framework analysis, discourse analysis, interpretative phenomenological analysis, content analysis, or another approach whose procedures and assumptions are unfamiliar to you.

Do not translate these into a generic idea that the researchers simply "looked for themes."

Ask how the analytical approach is defined, how data were prepared and examined, how codes, categories, themes, interpretations, or theoretical constructs were developed, who participated in the analytical process, and what procedures were used to support the credibility or transparency of the analysis where relevant to that methodological tradition.

Just as with quantitative methods, the appropriate appraisal criteria depend on the method. Avoid imposing standards from an unrelated research tradition simply because they are more familiar to you.

For computational or algorithmic methods, identify inputs, transformations, and outputs

Computational analyses can become opaque when a paper names algorithms, preprocessing pipelines, models, or software without explaining them in terms accessible to a non-specialist.

A useful first reconstruction is:

What went in → what happened to it → what came out → how was that output evaluated?

What data entered the procedure? Were observations excluded, normalized, transformed, tokenized, filtered, imputed, aggregated, or otherwise processed? What model or algorithm was then applied? What output did it produce? Against what reference or criterion was performance evaluated?

This does not provide a complete methodological appraisal, but it gives you a conceptual pipeline. You can then investigate the particular transformation or model that matters to your interpretation.

Ask whether you understand the assumptions that could change the conclusion

Methods make assumptions. Some are mathematical, some methodological, and some conceptual.

A statistical model may assume a particular relationship among variables. A measurement instrument may assume that certain observed responses represent an underlying construct. A qualitative approach may rely on particular epistemological commitments. An experimental design may assume that the comparison condition provides an appropriate counterfactual. A computational model may depend on training data and evaluation choices that shape its performance.

You do not necessarily need to catalogue every assumption. Focus on those whose violation or misinterpretation could materially change the conclusion you intend to draw.

If you do not know what the consequential assumptions are, that is itself a useful question to take to a methodological source or expert.

Separate “I do not understand it” from “the authors did it incorrectly”

Difficulty understanding a method is not evidence that the method is inappropriate.

This matters particularly when reading research outside your own discipline. A technique that appears unusual from your methodological background may be standard and well justified for the question being addressed.

Try to reach the first level before making the second judgment.

Also separate poor understanding from poor reporting

Sometimes the problem really is the paper.

An essential procedure may be described inadequately. A model may be named without enough information to reconstruct its specification. An instrument may be mentioned without explaining how it was scored. Important analytical decisions may be absent from both the main article and supplementary material.

Critical appraisal depends on adequate reporting. Young and Solomon emphasize that evaluating the appropriateness of the study design, key methodological features, statistical methods, and their interpretation is central to assessing research validity and usefulness.

If the information necessary for that evaluation is missing, do not invent it.

Your conclusion may simply be: the report does not provide enough information for me to determine this.

That is more defensible than assuming that standard or appropriate procedures were followed.

Ask for help when the uncertainty is consequential

There is a point at which independent reading becomes inefficient or unreliable.

If an unfamiliar analysis is central to a paper you intend to cite heavily, adapt, replicate, critique, or use for an important research decision, consultation may be warranted. A statistician, qualitative methodologist, disciplinary specialist, data scientist, measurement expert, or other appropriately qualified colleague may see an issue that is difficult to recognize from introductory explanations alone.

Ask a specific question rather than handing someone the paper and saying that you do not understand it.

For example: "The authors report an adjusted odds ratio from this model. I understand the outcome and predictors, but I am unsure what the adjustment changes about the interpretation. Can you help me check whether I am reading it correctly?"

Specific questions make expert help much more useful and expose exactly where your current understanding ends.

AI can help explain, but it should not become the authority

An AI tool can sometimes help translate technical terminology, break an equation into components, explain the broad purpose of a statistical procedure, or generate questions you should ask about an unfamiliar method.

That can make it useful as an initial explanatory aid.

But methodological explanations can be subtly wrong, oversimplified, or inappropriate to the exact implementation in the paper. AI-generated citations and references also require verification.

If the method matters to your interpretation, verify the explanation against the paper itself and authoritative methodological sources. When the stakes are higher, consult relevant expertise.

Use AI to help identify what you need to understand, not to outsource the methodological judgment you are trying to develop.

You may legitimately stop before achieving complete understanding

Research methods can lead downward indefinitely. Understanding one statistical model may lead to estimation theory, probability distributions, optimization, assumptions, diagnostics, and a considerable amount of mathematics. A qualitative methodology may lead into an extensive philosophical and methodological literature.

You need a stopping rule.

For ordinary critical reading, you may have reached sufficient understanding when you can explain:

• what the method was used to do;
• why it broadly fits the research question or analytical task;
• what the relevant output means;
• which important assumption or limitation affects your interpretation; and
• what conclusion the method does and does not support.

If you are going to perform the method yourself, review it formally, or make a methodological contribution involving it, that stopping point moves considerably further away.

The standard is not "Do I now understand everything?" It is "Have I resolved the uncertainty that matters to what I am about to claim or do?"

From “I Don’t Understand This Model” to a Usable Interpretation

The important move was narrowing the problem. You did not need to "learn multilevel statistics." You needed enough understanding of a particular model, in a particular study, for a particular interpretive purpose.

Common Mistakes When Methods or Analysis Become Difficult

Treat Confusion as a Question You Can Narrow

When you reach an unfamiliar method, do not choose immediately between skipping it and mastering it. Diagnose the gap first.

A useful final question is: If my understanding of this method is wrong, could it materially change what I am about to say about the paper?

If yes, you are not finished with that part of the paper yet.

When You Encounter a Method You Do Not Understand

Questions About Difficult Research Methods and Analyses

You Do Not Need Complete Mastery, but You Need Enough to Know What the Evidence Means

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.