Shodhyantri Engineering Services
Research · Analysis · Editing · Publication Support
Statistical support for the pooling stage of a meta-analysis: effect size conversion, fixed or random effects models, heterogeneity, sensitivity checks and publication bias.
Tell us about your meta-analysis statistics need. We reply within one working day.
Meta-analysis statistics support focuses on the quantitative core of a meta-analysis: turning results from many studies into comparable effect sizes and pooling them with the right model. Your search, screening and study selection decide what goes in. The statistics decide whether the pooled answer is believable. Errors in effect size conversion, model choice or heterogeneity reporting are among the most common reasons reviewers question a meta-analysis.
We work with the data you extract from included studies. We help you choose an effect size metric, convert reported statistics, run fixed or random effects models, explore heterogeneity through subgroups and meta-regression, test for small-study effects and produce forest and funnel plots. You remain the reviewer of the evidence and the author of the interpretation. The judgement on what the evidence means stays yours.
Each included study contributes an effect size and its variance. Common metrics are standardised mean difference such as Hedges' g for continuous outcomes, odds ratio or risk ratio for binary outcomes, and correlation coefficients transformed with Fisher's z. Studies are then weighted, usually by inverse variance, and pooled.
A fixed-effect model assumes one true effect shared by all studies. A random-effects model assumes true effects vary, and estimates both the mean effect and between-study variance, often written as tau-squared. Heterogeneity is described with Q, I-squared and prediction intervals. Publication bias is explored with funnel plots, Egger's regression and trim-and-fill, interpreted with caution.
Means, standard deviations, t values, F values, odds and correlations are converted carefully, with every formula recorded so reviewers can check your work.
We help you justify random or fixed effects using your research question and study diversity, not just whichever gives the cleaner result.
Subgroup analysis and meta-regression are planned around moderators named in advance, so explanations for variation are credible rather than opportunistic.
Funnel plots, Egger's test and trim-and-fill are run and interpreted cautiously, with their known limits stated clearly in your manuscript.
Analysis is done in R with metafor or meta, or in JASP or STATA, with scripts you can share as supplementary material.
Leave-one-out and influence analyses show whether a single study drives your pooled estimate, which reviewers almost always ask about.
Ten stages, each with a clear output. Nothing moves to the next stage until you have reviewed the last one.
— You review every stage first, then decide —
A sample of extracted values is rechecked against original papers to catch transcription and unit mistakes.
Effect signs are aligned so a positive value means the same thing across every included study.
Multiple effects from one sample are identified and handled to avoid double counting participants.
The pooling model and estimator for tau-squared are reviewed against study number and diversity.
Subgroups and meta-regression are limited to pre-specified moderators with enough studies to analyse.
Conclusions are checked against heterogeneity, risk of bias and bias test results before reporting.
Every stage is shared with notes explaining what was changed and why. You and your supervisor check it, and the next stage starts only after your go-ahead.
| Common problem | How Shodhyantri helps |
|---|---|
| Pooling studies with very different designs, populations or outcomes into one estimate. | Group studies by comparable design and outcome, pool only where combination makes sense, and narrate the rest separately. |
| Choosing a fixed-effect model only because random effects gave a wider interval. | Select the model from the research question and expected diversity of studies, and state the reasoning in your methods. |
| Reporting I-squared alone as if it measured how large the variation is. | Report tau-squared and a prediction interval along with I-squared, so readers see the actual spread of true effects. |
| Using several outcomes from the same study as if they were independent. | Select one effect per study using a stated rule, average them, or use multilevel or robust variance methods for dependence. |
| Running Egger's test with very few studies and treating its result as conclusive. | Interpret small-study tests cautiously when studies are few, and describe the result as limited evidence rather than proof. |
| Running many unplanned subgroup analyses and reporting only those that were significant. | Pre-specify moderators in the protocol, report all planned analyses, and label any additional ones clearly as exploratory. |
Keep the formula and source values for each converted effect size. Reviewers often request this and it saves time during revision.
Hedges' g corrects the small-sample bias in Cohen's d, which matters when several included studies have modest sample sizes.
When a paper lacks standard deviations or sample sizes, a short email to the corresponding author can rescue a usable study.
A prediction interval shows where the effect of a new study might fall and gives a more honest picture than confidence intervals alone.
Run leave-one-out analysis before drafting results. If one study drives everything, the discussion needs to say so plainly.
Follow PRISMA reporting items for the statistical sections, including effect measures, synthesis methods and certainty assessment.
Order studies sensibly, label axes with the direction of benefit and show weights, so readers can interpret plots quickly.
Pooled significance is not the same as certainty of evidence. Consider risk of bias and consistency, for example using GRADE.
Your work is handled by people who read your field's literature, not by a single generalist.
A PhD-qualified specialist from your field who checks the research logic, gap and contribution.
Handles statistics, models and qualitative coding, and documents every analytical decision.
Improves structure, flow, clarity and academic tone without changing your meaning.
Builds and validates MATLAB, Python, ANSYS or COMSOL work where your study needs it.
Checks grammar, consistency, numbering, citations and the small errors examiners notice.
Applies your university or journal template, reference style and front matter exactly.
No. Systematic review support covers protocol, search, screening and risk of bias. This page covers the statistics after studies are selected: effect sizes, pooling, heterogeneity, sensitivity and publication bias. Many scholars use both, but each can be taken separately depending on where you are.
Pooling is technically possible with two studies, but estimates of heterogeneity become unreliable with few studies, and tests for publication bias usually need around ten or more. With very few studies, a careful narrative synthesis may be more honest. We help you judge this.
We mostly use R with the metafor or meta packages because they are flexible and reproducible. JASP offers a friendly interface for simpler analyses, and STATA has good meta-analysis commands. We can work in whichever your supervisor or journal prefers.
That is very common. We convert means and standard deviations, t or F statistics, p-values with sample sizes, odds and correlations into a common metric using established formulas. Where conversion is not possible, we discuss whether to contact authors or exclude the study.
No. Publication decisions belong to journals and their peer reviewers. We help make the statistical work correct, transparent and well reported, which addresses frequent reviewer concerns, but acceptance also depends on your question, search quality and the journal's priorities.
Network meta-analysis, which compares several treatments at once, needs careful checks of transitivity and consistency. We can discuss whether your evidence base supports it and advise on approach. For most thesis projects, pairwise meta-analysis is the more practical choice.
There is no good or bad value in itself. I-squared shows the share of variation due to true differences rather than chance. Rough labels like low, moderate and high exist, but report tau-squared and a prediction interval too, and explore reasons for variation.
We provide research guidance, statistical analysis, editing, formatting and training support. All research content remains the scholar's own work, in line with UGC and university academic integrity norms. We do not guarantee publication, acceptance, indexing or academic outcomes: journals decide through peer review and universities award degrees.
Share your topic, draft or data. We will tell you honestly what it needs, and what it does not, before you pay anything.