RESEARCH SUPPORT
Meta-analysis data extraction support
for effect sizes
Effect sizes read from published reports, converted to one common metric with their variances and coded by a rule fixed in advance, so that each study enters the pooled analysis in the form its design allows, with each value traceable to its source.
No Commitment Required
Strictly Confidential
In-House PhD-Level Statisticians
Fixed Project Fee
Same-Day Response
What Does Extracting Effect Sizes From Studies Involve?
Meta-analysis data extraction support is research-analysis support for the stage before pooling: getting one comparable effect size and variance out of each published study. Most of the work is judgment applied consistently, such as which outcome and time point to take, how a median, a risk ratio or a test statistic becomes a common metric, and what to do when a paper reports too little. Missing standard deviations in meta-analysis are recovered by a stated method where one exists, and each converted value is flagged for later sensitivity checks. OnlineSPSS is a specialist statistical consultancy for academic, clinical and professional research, and its in-house PhD-level statisticians carry out the extraction. The review question and the inclusion decisions stay with the research team.
WHEN TO ENGAGE
Included Studies Report Their Results Differently
Once the included studies are collected, the way each one reports its outcome decides how much extraction work lies ahead.
Mixed Metrics Across Reports The studies give means and standard deviations, event counts, correlations or only test statistics, and one comparable effect size must be chosen and justified before any pooling.
Several Results Per Study A trial reports multiple outcomes, time points and subgroups, so the rule for which result is extracted must be fixed before coding, not decided study by study after the values are seen.
Gaps in the Reports A paper omits a standard deviation or gives a median with a range, and the missing value has to be obtained, estimated by a stated method or left out with the reason on record.
Two Extractors, Two Answers Two reviewers extracted the same studies and their values differ, so each discrepancy is settled against the source report and the level of agreement is documented.
WHAT IS EXTRACTED
Rules, Metrics, Variances and the Handover File
Every study passes through the same sequence, from the choice of metric to the dataset that goes forward for pooling.
Extraction Rules
Which outcome, time point and comparison is taken from each report is set in a coding rule before extraction, so two extractors take the same number for the model in Meta-Analysis.
Choice of Metric
The metric follows the outcome: standardized mean differences for continuous scales, risk or odds ratios for events and correlations for association, with the reason for each choice recorded.
Conversion and Variances
Reported statistics are converted to the chosen metric together with the sampling variance, and results rebuilt from a p-value or a figure are marked as less precise than reported ones.
Dependent Estimates
Estimates that share participants, such as two arms compared with one control group, are identified in the file so that the model can allow for the dependence and not count them as independent.
Moderator Coding
Study features that may explain differences between results are coded in advance for Meta-Regression, with a note of which codes rest on interpretation.
Handover to Pooling
The finished file goes forward to the model, with heterogeneity and publication bias checks, under Meta-Analysis & Evidence Synthesis, where the overall estimate is produced.
EXTRACTION PITFALLS
Errors Introduced Before Any Model Is Fitted
A pooled estimate is only as sound as the numbers entered, and these errors pass unnoticed because the software accepts any value it is given.
Standard Error Entered as SD A standard error typed where a standard deviation is expected gives the study a variance far too small and therefore too much weight, with no warning anywhere in the output.
Direction Not Aligned Scales on which a higher score is better and scales on which it is worse are pooled without reversing one, so real effects of opposite sign cancel each other out.
Adjusted Mixed With Raw Unadjusted and covariate-adjusted estimates are pooled as if they answered the same question, though they do not, so one type is chosen and applied throughout the file.
One Trial, Counted Twice Overlapping publications of a single trial, or several arms sharing one control group, enter as independent studies and make the pooled interval look narrower than it is.
Odds Ratios Read as Risks An odds ratio from a common outcome is read as a risk ratio, which overstates the effect when the outcome is not rare, so the metric is named in every table.
Converted Values Unflagged Values rebuilt from a median, a p-value or a figure are less exact than reported ones, and without a flag no sensitivity check can show how far they drive the result.
DELIVERABLES
The Extraction Record and the File It Yields
A reader auditing the review can follow any number in the pooled analysis back to the page it was taken from, using the documents below.
Extraction Form and Rules The coding form with a definition for every field, the rules for choosing among outcomes and time points, and the metric selected for each outcome, each with its reason.
Conversion Log Every converted value with its source page or table, the conversion rule applied and a flag marking it as calculated, estimated or taken as reported by the authors.
Pooling-Ready Dataset One row per effect with the estimate, its variance, the direction, the study identifier, the moderator codes and the markers for participants shared between comparisons.
Missing-Information List The studies with absent or unclear values, how each gap was handled, and which handling can be varied later in a sensitivity check of the pooled result.
Extractor Agreement Record Where two extractors disagreed, the values on both sides, the passage in the report that settled each and an agreement statistic suited to the type of coding.
Extraction Methods Note Statistical wording for the methods and results, matched to the output, for you to check and incorporate, describing the extraction and conversion steps and what was assumed where reports were incomplete.
ILLUSTRATIVE EXAMPLE
Extraction for a Review of School Asthma Programs
An illustration of how outcomes reported in unlike forms are brought to one metric before a review pools them.
ILLUSTRATIVE EXAMPLE · PEDIATRIC ASTHMA EDUCATION REVIEW RESEARCH · UNITED STATES
A doctoral researcher in nursing was reviewing randomized trials of school-based asthma education and needed one effect size per trial for symptom-free days before the trials could be pooled.
The trials reported symptom-free days as means with standard deviations, as medians with ranges, or as the share of children with controlled asthma, so the researcher and the statistician first fixed which outcome counted and set a rule for the follow-up visit nearest the end of the school term. Medians and ranges were converted by a stated approximation and flagged, and the proportions were kept as a separate outcome and not forced into the same metric. The complication was that two trials compared two education formats against one shared control group, which would have counted the control children twice, so the dependence was recorded in the file. Every estimate rebuilt from an approximation was marked for a later sensitivity check.
Outcome: The researcher received the extraction form, the conversion log, the pooling-ready dataset with each estimate and its variance, the flags for approximated values and the extraction methods note.
This example does not describe a real client or project. We never share details of actual client work, in line with the confidentiality we promise every researcher. This illustrative example is included to show the type, scope, and standard of work involved in this kind of analysis.
SOFTWARE
Tools for Converting and Pooling Effect Sizes
Conversion and pooling can run in a general statistical package, or the finished file can go into dedicated review software such as Comprehensive Meta-Analysis (CMA).
HOW WE ASSESS YOUR PROJECT
We Review Your Actual Research
Your assessment is based on a review of your actual research materials and analytical requirements, not a generic project category.
You Submit Your Project Details
Share whatever is available at the current stage, such as the research aim, protocol or proposal, dataset, variable list, manuscript or statistical requirements.
We Review Your Research and Define the Scope
Your statistically trained Research Coordinator reviews the materials to understand the study design, research questions, data structure, analytical requirements and expected deliverables. If clarification is needed, additional documents or information can be requested before the assessment is finalized.
Your Analysis Is Completed and Delivered
Once the scope and fixed project fee are agreed, the in-house statistician completes the work and the Research Coordinator remains the communication point through delivery, with the analysis, deliverables and timeline as agreed.
RESEARCH RESPONSIBILITY
The Review and Its Conclusions Are Yours
Which studies belong in a review is a judgment made against its protocol, and the extraction follows that judgment without altering it.
Agreed Statistical Support OnlineSPSS provides the agreed methodological, analytical and statistical reporting support, defined in a scope that is confirmed before any work begins.
No Intellectual Contribution OnlineSPSS does not claim any intellectual contribution to your review. Authorship of the review and its report remains yours.
Your Research Decisions The review question, the eligibility criteria, the choice of included studies, the interpretation of the pooled findings and the conclusions remain yours, with overall responsibility.
Confidential Handling Research data and project details are handled confidentially, project information is not published, and an NDA is available where required.
FREQUENTLY ASKED QUESTIONS
Extracting Effect Sizes From Published Studies
Medians in place of means, absent standard deviations and trials with several outcomes call for decisions that a protocol should settle.
What does meta-analysis effect size extraction cover?
Meta-analysis effect size extraction covers reading the results of each included study, choosing a common metric, converting the reported statistics to it with a variance, and coding the study features needed for pooling. Randomized, observational and diagnostic studies can all be included when the metric suits the design, and every value stays traceable to its source report.
Can a standard deviation be estimated from a median and a range?
Sometimes. Published methods estimate a mean and standard deviation from a median with a range or quartiles, and they work best when the underlying values are roughly symmetric and the sample is not small. The estimate is labeled approximate in the dataset, and a sensitivity analysis shows whether the pooled result changes when those studies are set aside.
Which result is extracted from a trial with several outcomes or time points?
One result per study is chosen by a rule fixed before extraction, such as the primary outcome named in the protocol or the follow-up closest to a set time. Choosing the largest effect after looking is a form of selective reporting. Where several results are meant to be kept, the model must allow for the correlation among them.
Does the extraction include searching for or screening studies?
No. The review team makes the search, screening and eligibility decisions against its protocol, and the extraction works on the studies you have included. Any other part of the review that you would like supported is settled in the assessment.
How do you protect my research data and confidentiality?
Your data and project details are handled confidentially, and project information is not published. Researcher names and institutions are not used publicly without permission. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, such as the protocol, the list of included studies and any extraction already done, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
One Coding Rule for
Every Included Study
Outcomes, time points and approximations settled in a coding rule at the outset keep two extractors consistent and leave a record that a peer reviewer can audit study by study.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014