Question & Search
Frame your PICO, PECO or PCC question and build database-ready search strings, with one-click links to PubMed and other databases.
What this does
Generates database-specific search strings from structured PICO terms using Boolean operators and MeSH/Emtree mappings.
When to use
At the start of a systematic review, after defining your research question.
How to use
- Enter search terms for each PICO component
- Add synonyms and MeSH terms
- Select target databases (PubMed, Embase, etc.)
- Copy the generated search string
Interpreting results
Each generated string combines your terms with AND/OR logic appropriate for the target database. Test the string in the actual database and refine as needed.
Key terms
- Boolean operators: AND narrows a search (both terms must appear), OR widens it (either term), NOT excludes a term.
- MeSH (Medical Subject Headings): PubMed's controlled vocabulary of standard topic tags. Emtree is the Embase equivalent. Searching these catches papers regardless of the exact words an author used.
Protocol
Draft and download your review protocol: PRISMA-P for systematic reviews, or a scoping-review protocol. Items prefill from your question and search where available.
What this does
Builds a structured review protocol from the official reporting items, prefilled from your Research Context, and downloads it as an editable Word document you can extend.
When to use
Before screening. A registered protocol (e.g. PROSPERO) reduces bias and is expected by most journals.
How to use
- Pick systematic review (PRISMA-P) or scoping review
- Prefill from your project, then complete each item
- Download the Word file and register it before screening
Full-text review
For studies included at title and abstract: record whether the full text was retrieved and whether it meets eligibility. Feeds the PRISMA flow.
What this does
The second screening stage. It takes the studies you included at title/abstract, tracks full-text retrieval, and records final eligibility with reasons for exclusion.
When to use
After title/abstract screening, before data extraction.
How to use
- For each study, mark whether the full text was Retrieved or Not retrieved
- For retrieved reports, mark Eligible or Exclude (with a reason)
- Sync to PRISMA, then send the eligible studies to extraction
PICO Framework
Structure your clinical research question using PICO. Choose PECO for observational/epidemiological studies, or PCC for scoping reviews.
What this does
Structures your research question into Population, Intervention/Exposure, Comparison, and Outcome components.
When to use
Before starting a systematic review, to define a focused and answerable research question.
How to use
- Select framework variant (PICO, PECO, or PCC)
- Fill in each component
- Use the structured question to guide your search strategy and eligibility criteria
Interpreting results
A well-formed PICO question drives every downstream step: search terms, inclusion criteria, data extraction fields, and the structure of your evidence synthesis.
Literature Screening
Upload references from any format, deduplicate, screen titles and abstracts with AI assistance, track PRISMA counts, and send included studies directly to extraction.
What this does
Import references from database exports, remove duplicates, and screen titles/abstracts against your eligibility criteria. AI scoring ranks references by relevance.
When to use
After running your database searches and exporting results as RIS, BibTeX, CSV, or PubMed XML.
How to use
- Upload reference files (RIS, BibTeX, CSV, XML supported)
- Set eligibility criteria (PICO, inclusion, exclusion)
- Use AI scoring to prioritize screening
- Mark each reference as include, exclude, or unclear
- Send included references to extraction
Interpreting results
AI scores (0-100) indicate estimated relevance. Always verify AI recommendations manually. PRISMA counts update automatically as you screen.
Key terms
- RIS, BibTeX, nbib: citation file formats that databases (PubMed, Embase, Scopus) and reference managers (EndNote, Zotero, Mendeley) export. Upload whichever your search produced.
- Deduplicate: removes the same reference found in more than one database.
- AI relevance score (0 to 100): an estimate of how well a reference matches your criteria, used to order screening. Always confirm by reading.
Eligibility Criteria & AI Settings
Data Extraction
What this does
AI-powered data extraction from published studies.
When to use
After collecting full-text PDFs of included studies.
How to use
- Pick a review type to load a field preset
- Add or remove fields so the AI extracts exactly what you need
- Upload one or more PDFs and extract
- Review, edit, then export or send to the characteristics table
Interpreting results
Extracted values should always be verified against the source. AI extraction is an aid, not a replacement for manual review.
Scoping Review
Map evidence on a topic, identify gaps, and describe available research.
11 fieldsSystematic Review
Critically appraise studies with risk of bias, quality scores, and inclusion criteria.
16 fieldsMeta-Analysis (Full)
All fields including effect sizes, CIs, statistical methods, confounders, risk of bias.
24 fieldsQuick Meta Extraction
Concise extraction focused on effect sizes, CIs, p-values - ready to paste into the meta-analysis tool.
15 fieldsDrop PDF files here
Browse your computer · Multiple files supported
Meta-analysis
Run pairwise meta-analysis with forest plots, funnel plots, and heterogeneity statistics.
What this does
Inverse-variance meta-analysis with fixed or random effects models.
When to use
After extracting effect sizes (OR, RR, HR, MD, SMD) and confidence intervals from 2+ studies.
How to use
- Enter study name, effect size, and 95% CI
- Select effect type and model
- Run analysis to get pooled estimate, forest plot, funnel plot
- Check heterogeneity (I2, Q) and publication bias (Egger, Begg)
Interpreting results
I2 < 25% = low heterogeneity, 25-75% = moderate, >75% = high. Egger p < 0.10 suggests publication bias. Check if the pooled CI crosses the null (1 for ratios, 0 for differences).
Key terms
- Fixed vs random effects: fixed assumes every study estimates one true effect; random allows the true effect to vary between studies. Use random when the studies differ in design or population.
- Weight: how much each study counts. More precise (usually larger) studies get more weight (inverse-variance weighting).
- Heterogeneity: how much the results differ beyond chance. Q (Cochran's Q) is the test; I2 is the percent of variation from real differences (under 25 percent low, 25 to 75 moderate, over 75 high); tau2 (and its square root tau) is the estimated between-study variance.
- DL vs REML: two ways to estimate tau2. DerSimonian-Laird (DL) is the classic method; REML is often preferred for continuous outcomes. They rarely change the conclusion.
- z statistic: tests whether the pooled effect differs from no effect.
- Egger and Begg tests: screen for small-study effects and possible publication bias (a p below about 0.10 is a flag, not proof).
- Subgroup analysis and meta-regression: subgroup analysis pools studies within groups and tests whether the effect differs between them; meta-regression relates the effect to a study-level variable (a moderator, for example mean age).
1. Framework
2. Study design
3. Available data
4. Select outputs
| Author(s) | Year | Effect Size | 95% CI Lower | 95% CI Upper | Subgroup |
|---|
PRISMA 2020 Flow Diagram
Generate a publication-ready PRISMA 2020 flow diagram for your systematic review.
What this does
PRISMA 2020 flow diagram generator.
When to use
To report the study selection process in a systematic review.
How to use
- Enter counts for each stage
- Provide exclusion reasons
- Select PRISMA variant (2009 or 2020)
- Export as SVG or PNG
Interpreting results
The diagram shows how many records were found, screened, excluded, and included. Transparent reporting of the selection process.
Chart Style & Colours
Risk of Bias Assessment
Assess study quality using validated tools: RoB 2, NOS, and AXIS for multiple study designs.
What this does
Risk of bias assessment using standard tools (RoB 2, ROBINS-I, NOS, AXIS, JBI).
When to use
During quality assessment of included studies.
How to use
- Select the appropriate tool for your study design
- Add studies and rate each domain
- View the traffic light summary
- Export as CSV, SVG, or PNG
Interpreting results
Low = low risk. Some concerns = potential issues. High = likely bias affecting results. Overall judgement follows the worst domain rating.
Key terms
- Domain: a specific source of bias you judge (for example how participants were randomised, or how the outcome was measured).
- The tools: RoB 2 for randomised trials (5 domains); ROBINS-I for non-randomised studies of interventions (7 domains); NOS (Newcastle-Ottawa Scale) for cohort and case-control studies; AXIS for cross-sectional studies; JBI (Joanna Briggs Institute) checklists for several designs.
- Judgements: RoB 2 uses Low, Some concerns, High. ROBINS-I uses Low, Moderate, Serious, Critical. The overall rating follows the worst domain.
- Traffic light summary: the coloured grid of each study's rating per domain (green low, yellow some concerns, red high).
Leave-One-Out Sensitivity Analysis
Assess robustness by systematically removing one study at a time. Identifies influential studies.
What this does
Re-runs the meta-analysis k times, each time removing one study. Shows how each study influences the pooled estimate.
When to use
After running a meta-analysis, to check if the result depends on any single study.
How to use
- Run a meta-analysis first
- Open Leave-One-Out and click Run
- Review the table and forest plot showing pooled estimates with each study removed
Interpreting results
If removing one study changes the pooled estimate direction or significance, that study is influential. Consider subgroup analysis or investigate why it differs.
Key terms
- Leave-one-out: re-runs the pooled analysis with each study removed in turn, to see how much any single study drives the result.
- Influential: a study is flagged when removing it shifts the pooled estimate enough to matter (changing its direction or whether it stays statistically significant).
| Author(s) | Year | Effect Size | 95% CI Lower | 95% CI Upper |
|---|
Network Meta-Analysis
Visualise your treatment network, compute indirect comparisons, rank treatments by P-score, and generate a league table. Based on frequentist NMA methodology.
What this does
Network meta-analysis comparing multiple treatments simultaneously.
When to use
When 3+ treatments have been compared across studies, with both direct and indirect evidence.
How to use
- Add treatments
- Enter pairwise comparisons (studies, effect, SE)
- Generate network diagram
- Review league table, rankings, and P-scores
Interpreting results
P-score near 1 = likely best treatment. Check network connectivity and consistency. Inconsistency suggests the indirect evidence disagrees with direct evidence.
Key terms
- Direct vs indirect evidence: direct compares two treatments tested head to head; indirect compares them through a shared comparator (A vs B inferred from A vs C and B vs C).
- Bucher method: the calculation that combines direct comparisons into an indirect one.
- League table: a grid of every treatment against every other, read as row versus column.
- P-score: a 0 to 1 score for how likely each treatment is to be among the best (near 1 is best). It is the frequentist version of SUCRA.
- Consistency: agreement between the direct and indirect evidence in a closed loop of the network. Disagreement is called inconsistency.
- SE (standard error): the precision of a comparison's effect estimate, smaller is more precise.
Add treatments and click Generate
Direct comparisons (edges): m
Degrees of freedom for inconsistency: m − k + 1
Total studies: n
≈ Φ( μˆ_AB / SE(μˆ_AB) )
where μˆ_AB is the NMA estimate for A vs B
SE² = SE²_AB + SE²_CB
Inconsistency factor IF = θ_direct − θ_indirect
network <- mtc.network(data.re = your_data)
model <- mtc.model(network, type="consistency", linearModel="random")
results <- mtc.run(model, n.adapt=5000, n.iter=20000, thin=10)
summary(results)
rank.probability(results)
References: Rücker G. Stat Med. 2012;31:1236–1251. | Rücker G, Schwarzer G. BMC Med Res Methodol. 2015;15:58. | Salanti G et al. Ann Intern Med. 2011;154:544–553. | Dias S et al. Stat Med. 2013;32:739–776.
Forest Plot Builder
Build a publication-ready forest plot from any effect estimates - no raw data needed. Enter your studies manually or paste from a spreadsheet.
What this does
Standalone forest and funnel plot builder with full style control.
When to use
When you need publication-ready plots from pre-computed effect sizes.
How to use
- Enter study data (name, effect, CI, weight)
- Select effect type and adjust styles
- Download as SVG or PNG
Interpreting results
Forest plot: each line = one study CI. Diamond = pooled estimate. Funnel plot: asymmetry suggests publication bias.
| Study / Label | Effect | CI Lower | CI Upper | Weight % | Group |
|---|
Effect Size Calculator
Convert between effect measures, calculate NNT/NNH, and transform effect sizes for meta-analysis.
What this does
Three calculators: (1) NNT and absolute risk from an effect estimate plus a baseline risk; (2) convert between effect measures (OR, RR, RD, SMD); (3) recover the standard error and variance from a 95% confidence interval.
When to use
To translate an effect into plain numbers (how many people benefit), to convert a measure into the one your analysis needs, or to get the SE a meta-analysis requires from a published CI.
How to use
- Pick a tab and enter the values; results update as you type.
- For NNT, enter the effect and the baseline (control) risk.
- Use Sample on any tab for a worked example.
Key terms
- CER (control event rate): the outcome rate in the control or untreated group, i.e. the baseline risk.
- EER (experimental event rate): the outcome rate in the treated or exposed group.
- RR (risk ratio): EER divided by CER. Below 1 treatment lowers risk; above 1 it raises it.
- OR (odds ratio): the odds of the outcome in the treated group divided by the odds in the control group.
- HR (hazard ratio): the ratio of event rates over time, used in survival analysis.
- RD / ARR (risk difference, absolute risk reduction): CER minus EER, the absolute change in risk. A negative value is an absolute risk increase (ARI).
- RRR (relative risk reduction): the share of the baseline risk removed by treatment, equal to 1 minus RR.
- NNT (number needed to treat): how many people must be treated for one extra good outcome, equal to 1 divided by ARR. NNH (number needed to harm) is the same idea for one extra bad outcome.
- SMD (standardised mean difference) and Cohen's d: a mean difference in standard-deviation units (about 0.2 small, 0.5 medium, 0.8 large).
- SE (standard error) and 95% CI (confidence interval): how precise an estimate is.
Interpreting results
For ratios (OR, RR, HR) a value of 1 means no difference between groups. A smaller NNT means a larger benefit per person treated.
CONSORT Flow Diagram
Generate a CONSORT 2010 flow diagram for your randomised controlled trial.
What this does
Generates a CONSORT 2010-compliant flow diagram showing participant flow through enrollment, allocation, follow-up, and analysis.
When to use
When reporting results of a randomised controlled trial.
How to use
- Enter participant counts at each stage
- Specify exclusion and loss-to-follow-up reasons
- Export as SVG or PNG
Interpreting results
The diagram must account for every participant from enrollment to analysis. Discrepancies between allocated and analysed numbers should be explained.
Chart Style & Colours
GRADE Evidence Table
Rate the certainty of evidence for each outcome using the GRADE approach.
What this does
GRADE certainty of evidence assessment for each outcome.
When to use
After completing meta-analysis and risk of bias assessment.
How to use
- Add an outcome
- Set study design (RCT starts at High, Observational at Low)
- Rate 5 downgrade and 3 upgrade criteria
- Review the final certainty rating
Interpreting results
High = very confident. Moderate = moderately confident. Low = limited confidence. Very Low = very little confidence. Each serious downgrade drops one level.
Key terms
- Certainty of evidence: how much the estimate for an outcome can be trusted, rated High, Moderate, Low, or Very Low.
- Five reasons to downgrade: risk of bias (flaws in how studies were run), inconsistency (results differ across studies), indirectness (the evidence does not match your exact question, population, or outcome), imprecision (wide confidence intervals or few events), and publication bias (missing studies).
- Three reasons to upgrade (for observational evidence): a large effect, a dose-response gradient, and plausible confounding that would only work against the observed effect.
- RCTs start at High and observational studies at Low; each serious concern moves the rating one level down (or up).
Characteristics of Included Studies
Build the "Table 1" summary of included studies for your systematic or scoping review. Add studies as rows and define columns based on your review type.
What this does
Builds the characteristics table summarizing key attributes of each included study. Column presets match common review types.
When to use
After data extraction, to create the summary table required in every systematic review.
How to use
- Select a column preset (scoping, systematic, meta-analysis, RCT) or customize columns
- Add studies as rows
- Fill in each cell
- Export as CSV or copy to your manuscript
Interpreting results
This table provides readers with an overview of included studies. Ensure consistent reporting across all rows. Missing data should be marked explicitly.
Randomization Tool
Generate allocation sequences for clinical trials. Supports simple, block, and stratified randomization.
What this does
Generates randomization sequences for clinical trials using simple, block, or stratified methods with configurable allocation ratios.
When to use
Before starting participant enrollment in a clinical trial.
How to use
- Set the number of participants and groups
- Choose randomization method (simple, block, stratified)
- Set block size if using block randomization
- Generate and export the allocation sequence
Interpreting results
Block randomization ensures balanced group sizes at regular intervals. Stratified randomization ensures balance across important prognostic factors. Keep the sequence concealed from recruiters.
Trend Analysis & Segmented Regression
Joinpoint regression with automatic BIC-based model selection. Calculate APC (Annual Percent Change) and AAPC (Average Annual Percent Change) with Delta Method confidence intervals.
What this does
Joinpoint regression with BIC/AIC model selection. Calculates APC and AAPC.
When to use
For time-series data (e.g. annual incidence rates, mortality trends).
How to use
- Paste Year and Value columns (supports multiple outcome columns)
- Select scale (log for rates, linear for counts)
- Choose auto or manual joinpoint selection
- Review APC per segment and overall AAPC
Interpreting results
APC = Annual Percent Change within a segment. AAPC = weighted average across all segments. Negative APC = decreasing trend. Joinpoints mark where the trend direction changes.
Key terms
- Joinpoint: a year where the trend changes slope. The tool fits straight segments joined at these points.
- APC (annual percent change): the yearly percent change within one segment. AAPC (average APC): a weighted average across all segments.
- Log-linear vs linear scale: pick log-linear for rates, proportions, or percentages (a constant percent change), and linear for counts or values that change by a fixed amount.
- BIC / AIC / AICc: scores that choose how many joinpoints to keep, balancing fit against simplicity. Leave on the default unless you want to set the number yourself.
- Weighted least squares: the fitting method, giving more weight to more precise points; the AAPC confidence interval uses the delta method.
Year Rate 2005 8.2 2006 9.1 2007 10.5 ...
Year Incidence Mortality 2005 8.2 3.1 2006 9.1 3.4 ...
year,sex,onset,count,population 1990,Female,Early,241334,1081823027 1990,Male,Early,264921,1109228616 ...
Statistical Analysis
Comprehensive statistics center - descriptive, comparison, correlation, regression, survival, diagnostic accuracy, prediction validation, and causal methods. Paste data from Excel, upload CSV, or enter manually. All in-browser calculations use transparent, established formulas. Code templates available for R and Python.
What this does
Statistical tests: t-test, chi-square, ANOVA, correlation, regression, Kaplan-Meier, ROC.
When to use
For individual study analysis or exploratory data analysis.
How to use
- Select the test type
- Paste or enter your data
- Run the analysis
- Review test statistics and p-values
Interpreting results
p < 0.05 is the conventional threshold for statistical significance. Always consider effect size alongside p-values. Check assumptions (normality, independence) before interpreting.
Key terms by tab
Comparison: a t-test compares two means (Welch does not assume equal variances); ANOVA compares three or more means; Mann-Whitney, Wilcoxon signed-rank, and Kruskal-Wallis are rank-based versions for non-normal data; chi-square, Fisher's exact, and McNemar's test associations in count tables; p-value corrections (Bonferroni, Holm, Benjamini-Hochberg) adjust for testing many hypotheses.
Correlation: Pearson (linear) and Spearman or Kendall's tau (rank-based) measure association from -1 to 1. ICC (intraclass correlation) measures agreement between raters.
Regression: OLS for continuous outcomes; logistic gives an OR (odds ratio); log-binomial and modified Poisson give an RR (risk ratio); Poisson and negative binomial give an IRR (incidence rate ratio, for counts; negative binomial adds theta for overdispersion); Cox gives an HR (hazard ratio); ordinal and multinomial handle ordered or unordered categories. beta is the coefficient, SE its standard error, R2 the variance explained.
Survival: Kaplan-Meier estimates survival over time; censored means the event had not happened by last follow-up; the log-rank test compares groups.
Diagnostic and ROC: sensitivity (true positives found), specificity (true negatives found), PPV/NPV (predictive values), DOR (diagnostic odds ratio), AUC (area under the ROC curve, 0.5 is chance and 1.0 is perfect), and the Youden index (best cut-off).
Prediction: the confusion matrix and accuracy, precision, recall, F1 summarise classification; the Brier score measures probability accuracy (lower is better); calibration checks predicted against observed risk; k-fold cross-validation tests on held-out data.
Causal and clustering: the Mann-Kendall test detects a monotonic trend; k-means groups observations into clusters (WCSS is within-cluster spread, used by the elbow plot).
| Disease + | Disease − | |
|---|---|---|
| Test + | ||
| Test − |
AI Data Advisor
Upload your dataset and describe your research objectives. Our AI will analyse your data structure, suggest appropriate statistical analyses and figures, and generate a ready-to-use prompt you can take to any AI assistant (Claude, ChatGPT, etc.) to instantly get complete R and Python code - without having to explain your data yourself.
What this does
Analyses your dataset structure and research objectives. Suggests statistical tests, figures, and generates a complete prompt for R/Python code generation.
When to use
When you have a dataset and need guidance on which analyses to run, or want ready-made code for R or Python.
How to use
- Upload a CSV or Excel file
- Describe your research objectives
- Review suggested analyses and figures
- Copy the generated prompt into any AI tool for complete code
Interpreting results
Suggestions are based on variable types and research context. Always verify that suggested tests match your study design assumptions. The generated prompt includes your actual column names and data types.
Methods, Transparency & Validation
Complete documentation of every statistical formula, algorithm, and decision rule used by Kovex. Each section includes the mathematical specification, academic references, and downloadable R replication code so you can independently verify our results.
What this does
Documents every formula, algorithm, and decision rule used across all Kovex tools. Includes R replication code for independent verification.
When to use
When writing your methods section, responding to peer review, or verifying that Kovex calculations match your expectations.
How to use
- Browse by tool or statistical method
- Review the mathematical specification
- Download R code to replicate any calculation independently
- Cite the listed references in your manuscript
Interpreting results
This page is reference documentation. Use it to confirm which formulas Kovex applies and to write accurate methods descriptions for publication.
Meta-analysis - Pooled Effect & Heterogeneity
▼Kovex implements both fixed-effect (Inverse-Variance) and random-effects (DerSimonian-Laird) models for pooling study-level effect estimates.
where v_i = variance of study i effect estimate
C = Sum(w_i) - Sum(w_i^2) / Sum(w_i)
tau^2 = max( 0, (Q - (k-1)) / C )
SE(theta_hat) = sqrt(1 / Sum(w_i*))
95% CI: theta_hat +/- 1.96 * SE(theta_hat)
Back-transformed: exp(theta_hat), exp(CI bounds)
tau = sqrt(tau^2)
H^2 = Q / (k-1)
p-value: chi-squared distribution with (k-1) df
References: DerSimonian R, Laird N. Control Clin Trials. 1986;7:177-188. | Higgins JPT, Thompson SG. Stat Med. 2002;21:1539-1558. | Borenstein M et al. Introduction to Meta-Analysis. Wiley; 2009.
R Replication Code
# Replicate Kovex meta-analysis in R
library(meta)
# Example: generic inverse-variance input
study <- c("Study A","Study B","Study C","Study D","Study E")
effect <- c(0.42, 0.58, 0.31, 0.67, 0.45) # log(OR) or SMD
se <- c(0.18, 0.22, 0.15, 0.25, 0.20)
m <- metagen(TE = effect, seTE = se, studlab = study,
sm = "OR", method.tau = "DL",
fixed = TRUE, random = TRUE)
summary(m)
forest(m)
funnel(m)
Leave-one-out Sensitivity Analysis
▼Each study is systematically removed and the meta-analysis re-run on the remaining k-1 studies to identify influential studies.
theta_hat(-i) = pooled estimate omitting study i
tau^2(-i) = between-study variance omitting study i
I^2(-i) = heterogeneity omitting study i
Reference: Viechtbauer W, Cheung MWL. Res Synth Methods. 2010;1:112-125.
R Replication Code
# Leave-one-out in R library(metafor) yi <- c(0.42, 0.58, 0.31, 0.67, 0.45) sei <- c(0.18, 0.22, 0.15, 0.25, 0.20) res <- rma(yi = yi, sei = sei, method = "DL") loo <- leave1out(res) print(loo) forest(loo)
Effect Size Calculator - NNT, ARR, RR, OR, SMD
▼RR = EER / CER
OR = (a*d) / (b*c)
RD = EER - CER
ARR = CER - EER
NNT = 1 / |ARR|
SE(ln RR) = sqrt(1/a - 1/(a+b) + 1/c - 1/(c+d))
SE(RD) = sqrt(EER*(1-EER)/(a+b) + CER*(1-CER)/(c+d))
SD_pooled = sqrt(((n1-1)*SD1^2 + (n2-1)*SD2^2) / (n1+n2-2))
Hedges' g = d * (1 - 3/(4*(n1+n2)-9))
References: Borenstein M et al. Wiley; 2009. | Zhang J, Yu KF. JAMA. 1998;280:1690-1691.
R Replication Code
# Effect size calculations in R
library(epitools)
library(meta)
# 2x2 table: a=15, b=85, c=30, d=70
tab <- matrix(c(15, 85, 30, 70), nrow=2, byrow=TRUE)
oddsratio.wald(tab)
riskratio.wald(tab)
# SMD (Hedges' g)
metacont(n.e=50, mean.e=12.3, sd.e=3.1,
n.c=50, mean.c=10.1, sd.c=2.9, sm="SMD")
Funnel Plot & Publication Bias
▼y-axis: SE(theta_i) inverted
Pseudo-95% CI: theta_hat +/- 1.96 * SE
Asymmetry suggests publication bias but may also reflect heterogeneity or small-study effects. Formal tests require 10+ studies.
References: Sterne JAC et al. BMJ. 2011;343:d4002. | Egger M et al. BMJ. 1997;315:629-634.
Joinpoint Trend Analysis - APC & AAPC
▼Kovex fits piecewise log-linear regression to time-series rate data, identifying points where the trend changes direction (joinpoints).
where (x)+ = max(0, x) and tau_j are joinpoints
95% CI: 100 * (exp(slope +/- 1.96*SE) - 1)
Weights = segment length / total period
References: Kim HJ et al. Stat Med. 2000;19:335-351. | Muggeo VMR. Stat Med. 2003;22:3055-3071. | Clegg LX et al. Stat Med. 2009;28:3670-3682.
R Replication Code
# Joinpoint trend analysis in R
library(segmented)
year <- 2000:2020
rate <- c(12.1,12.4,12.8,13.5,14.2,14.8,15.1,15.0,14.6,14.2,
13.8,13.5,13.9,14.5,15.2,16.1,17.0,17.8,18.2,18.5,18.7)
fit <- lm(log(rate) ~ year)
seg <- segmented(fit, seg.Z = ~year, npsi = 2)
summary(seg)
slope(seg) # APC per segment
plot(seg)
Network Meta-Analysis - P-score & League Table
▼Kovex implements frequentist NMA with Bucher indirect comparisons, league table generation, node-splitting consistency assessment, and P-score ranking.
var(theta_AB) = var(theta_AC) + var(theta_BC)
where Phi() = standard normal CDF
References: Salanti G. Res Synth Methods. 2012;3:80-97. | Rucker G, Schwarzer G. Res Synth Methods. 2015;6:260-267. | Bucher HC et al. J Clin Epidemiol. 1997;50:683-691.
R Replication Code
# NMA in R using netmeta
library(netmeta)
treat1 <- c("A","A","B")
treat2 <- c("B","C","C")
TE <- c(0.5, 0.8, 0.3)
seTE <- c(0.2, 0.25, 0.22)
net <- netmeta(TE, seTE, treat1, treat2, sm="OR")
summary(net)
netrank(net) # P-scores
netheat(net) # Consistency
netleague(net) # League table
Statistics Suite - Tests, Correlation, Survival, ROC
▼df (Welch) = (s1^2/n1 + s2^2/n2)^2 / ((s1^2/n1)^2/(n1-1) + (s2^2/n2)^2/(n2-1))
E_ij = (row_total_i * col_total_j) / grand_total
df = (rows-1) * (cols-1)
t = r * sqrt((n-2)/(1-r^2)), df = n-2
SE(S) = S(t) * sqrt(Sum[ d_j / (n_j*(n_j - d_j)) ]) (Greenwood)
SE(AUC) via DeLong method
MS_between = SS_between / (k-1)
MS_within = SS_within / (N-k)
References: Kaplan EL, Meier P. J Am Stat Assoc. 1958;53:457-481. | DeLong ER et al. Biometrics. 1988;44:837-845. | Welch BL. Biometrika. 1947;34:28-35.
R Replication Code
# Statistics suite replication # t-test t.test(x, y, var.equal = FALSE) # Chi-squared chisq.test(matrix(c(a,b,c,d), nrow=2)) # Correlation cor.test(x, y, method = "pearson") # Kaplan-Meier library(survival) fit <- survfit(Surv(time, event) ~ group, data = df) summary(fit) # ROC / AUC library(pROC) roc_obj <- roc(label, prob) auc(roc_obj) ci.auc(roc_obj)
Risk of Bias Assessment Tools
▼Five domains: randomization process, deviations from interventions, missing outcome data, outcome measurement, selection of reported results. Each: Low / Some Concerns / High. Overall = most severe domain.
Seven domains: confounding, selection, classification of interventions, deviations, missing data, measurement, reported results. Judgement: Low / Moderate / Serious / Critical / No information.
Case-control: max 9 stars (Selection x4, Comparability x2, Exposure x3)
Thresholds: 7-9 = Good (low risk) | 5-6 = Fair (moderate) | 0-4 = Poor (high risk)
20 items across 5 sections (Introduction, Methods, Results, Discussion, Other). Each Yes/No/Unclear. Section summary scores presented as ratios.
13 items covering randomization, allocation concealment, blinding, completeness of follow-up, intention-to-treat analysis, outcome measurement, and statistics. Each Yes / No / Unclear / Not applicable.
References: Sterne JAC et al. BMJ. 2019;366:l4898. | Sterne JAC et al. BMJ. 2016;355:i4919. | Wells GA et al. Ottawa Hospital Research Institute. | Downes MJ et al. BMJ Open. 2016;6:e011458. | Barker TH et al. JBI Evid Synth. 2023;21(3):494-506.
Licensing: RoB 2, ROBINS-I and JBI are CC BY-NC-ND 4.0; AXIS is CC BY-NC 4.0; NOS has no formal license stated. Reproduced here with attribution for non-commercial research use. See each tool's page for the citation and license note.
GRADE Evidence Certainty
▼Downgrade (-1 or -2 each):
1. Risk of bias 2. Inconsistency 3. Indirectness 4. Imprecision 5. Publication bias
Upgrade (+1 each):
1. Large effect (RR >2 or <0.5) 2. Dose-response 3. Plausible confounders reduce effect
Final = max(1, min(4, start - downgrades + upgrades))
4=High 3=Moderate 2=Low 1=Very Low
Reference: Guyatt G et al. J Clin Epidemiol. 2011;64:383-394.
Randomization Algorithm
▼Uniform [0,1] = X(n+1) / 2^32
Within each block: Fisher-Yates shuffle of group labels
Guarantees balance after every b participants
For research planning and education only. NOT validated for registered clinical trials. Use IVRS/IWRS for GCP compliance.
Reference: Schulz KF, Grimes DA. Lancet. 2002;359:515-519.
PRISMA 2020 & CONSORT Flow Diagrams
▼Reports sought = Screened - Excluded (title/abstract)
Reports assessed = Sought - Not retrieved
Included = Assessed - Excluded (full-text)
Two-arm parallel RCT flow: enrollment, allocation, follow-up (lost/discontinued), analysis. All counts user-entered; SVG rendered with fixed coordinate layout.
References: Page MJ et al. BMJ. 2021;372:n71. | Schulz KF et al. BMJ. 2010;340:c332.
AI-Powered Features (Extraction, Screening, Advisor)
▼AI features use large language models (LLMs) via API. These are assistance tools, not autonomous decision-makers.
Extracts structured fields from PDFs with source quotes and confidence ratings (High/Medium/Low). All output must be independently verified by a trained reviewer.
Scores references against user-defined inclusion/exclusion criteria. Produces include/exclude recommendation with reasoning. Final decisions must be made by human reviewers.
Analyses uploaded datasets and recommends appropriate statistical methods. Generates R and Python code snippets. Users must verify appropriateness for their specific research question.
AI outputs may contain errors or hallucinations. All AI-generated content must be reviewed and verified by a qualified researcher before use in any publication.
Figure Studio (Publication Figures)
▼Figure Studio renders publication figures in-browser from pasted or piped data, with control over labels, fonts, palette, height, gridlines, and a log x-axis for ratio measures. Each figure type ships a sample dataset and a downloadable template, and meta-analysis results can be sent straight to it.
Two views for showing how an exposure effect varies across strata of a third variable. The interaction plot draws the outcome across levels of the modifier, one line per exposure group: non-parallel lines indicate effect modification. The subgroup forest plots the effect estimate and 95% CI within each stratum against a null reference line, with the overall/pooled row drawn as a diamond.
Forest plots size markers by weight and mark the pooled estimate as a diamond. Charts export as PNG and SVG. Rendering uses Plotly.js (MIT licensed).
References: VanderWeele TJ, Knol MJ. A tutorial on interaction. Epidemiol Methods. 2014;3(1):33-72. | Plotly Technologies Inc. Collaborative data science. Montréal, QC; 2015.
Legal & Privacy
Terms of use, privacy policy, data handling practices, and liability information for the Kovex research platform.
What this does
Displays terms of use, privacy policy, and data handling practices for the platform.
When to use
Before using the platform, or when your institution requires documentation of data handling for ethics review.
How to use
- Read the terms and privacy policy
- Statistical tools run in your browser; AI features, the PubMed search, and (if you sign in) account and project sync send specific data off your device (see the Privacy Policy below)
Interpreting results
This is a reference page. No calculations or outputs are produced.
IMPORTANT: Kovex is a research assistance tool for use by trained researchers. It is NOT a medical device, clinical decision support system, or substitute for professional judgement. All outputs must be independently verified before use in any publication, guideline, or clinical decision.
Terms of Use
1. Acceptance
By accessing and using Kovex (“the Platform”), you accept and agree to be bound by these Terms. If you do not agree, do not use the Platform.
2. Permitted Use
Kovex is intended solely for:
- Academic and scientific research purposes
- Systematic review, scoping review, and meta-analysis methodology support
- Research education and training
- Generating figures and statistical outputs intended to support scholarly publication
3. Prohibited Use
You may not use Kovex for:
- Clinical diagnosis, treatment decisions, or patient care of any kind
- Regulatory submissions without independent validation by a qualified statistician
- Any purpose requiring medically validated or CE/FDA-cleared software
- Registered clinical trial randomization (use certified IVRS/IWRS for GCP compliance)
- Uploading documents containing patient-identifiable or sensitive personal data
4. User Responsibilities
- You are solely responsible for verifying all outputs before use in any publication or research document
- You should independently verify critical analyses before publication using established software (e.g., RevMan, R, Stata)
- You must comply with the terms of any third-party tools or methodologies implemented in the Platform
- You acknowledge that Kovex provides downloadable R replication code for every computational tool (see Methods) to facilitate independent verification
5. Your Content
You retain full ownership of any research data, uploaded documents, or other content you submit to the Platform (“Your Content”). By using the Platform you grant Kovex a limited, non-exclusive licence to process Your Content solely to deliver the services you request (e.g., cloud sync, AI feature processing). You represent and warrant that you have the right to submit Your Content and that doing so does not violate any law, regulation, or third-party right.
6. Service Availability
Kovex is provided on an “as available” basis. We do not guarantee uninterrupted access and may modify, suspend, or discontinue features at any time. AI-powered features depend on third-party LLM providers and may be subject to their availability and rate limits.
7. Subscriptions & Billing
- The core research tools are free. Paid plans (Pro, Team, Institution) add managed AI, collaboration, and administrative features as described on the Plans page.
- Paid plans are billed monthly or annually through our merchant of record (Lemon Squeezy / Stripe), who is the seller of record and handles payment and applicable taxes.
- Managed-AI allowances are per the plan; usage beyond the allowance requires credits or your own API key. If you connect your own key, that AI usage is billed to you by your provider.
- You can cancel at any time; access continues until the end of the paid period. Refunds follow the merchant of record's policy and applicable consumer law.
- Prices may change with notice; changes do not affect the current paid period.
Disclaimers & Limitation of Liability
No Medical Advice
Kovex does not provide medical, clinical, regulatory, or professional advice of any kind. Nothing on this Platform constitutes or should be construed as medical advice. Always consult qualified professionals for clinical decisions.
Research Tool Disclaimer
All tools - including statistical calculators, risk of bias assessments, GRADE ratings, PRISMA diagrams, and AI-extracted data - are provided as research aids to assist trained researchers. They are not validated for direct use in clinical practice or regulatory processes without independent verification.
AI Output Disclaimer
Artificial intelligence outputs (including data extraction, summarisation, screening scores, and any AI-assisted features) may contain errors, omissions, or hallucinations. All AI-generated content must be reviewed and verified by a qualified human researcher before use in any publication, guideline, or decision. AI features use large language models (LLMs) whose behaviour may change over time as providers update their models.
Calculation Accuracy & Transparency
Kovex implements established published statistical methods with full formula documentation available on the Methods page. Downloadable R replication code is provided for every computational tool so users can independently verify results. Despite these transparency measures, users should always cross-validate calculations before publication. Kovex accepts no liability for errors in user-entered data or for misinterpretation of outputs.
Limitation of Liability
To the fullest extent permitted by law, Kovex and its developers shall not be liable for any direct, indirect, incidental, special, or consequential damages arising from use of the Platform, including but not limited to loss of data, business interruption, research errors, or reliance on Platform outputs.
No Warranty
The Platform is provided “as is” and “as available” without warranty of any kind, express or implied, including warranties of merchantability, fitness for a particular purpose, or non-infringement.
Privacy Policy
Last updated: June 2026 · Version 1.3
This Privacy Policy explains how Kovex Labs (“Kovex”, “we”, “us”, “our”) collects, uses, stores, and protects your personal data when you use the Kovex platform at covexe.com. It applies to all users worldwide and specifically addresses your rights under the European Union General Data Protection Regulation (GDPR), the Taiwan Personal Data Protection Act (Taiwan PDPA), and the California Consumer Privacy Act (CCPA).
By creating an account or using the Platform, you acknowledge that you have read this Privacy Policy. If you do not agree, please discontinue use.
1. Data Controller
The data controller responsible for your personal data is:
Kovex Labs is an independent software project operated by Abdiwahab M. Ali, Taiwan.
Platform: covexe.com
Contact: help@covexe.com
2. What Personal Data We Collect and Why
| Category | Data collected | Purpose | Legal basis (GDPR) |
|---|---|---|---|
| Account | Email address, display name | Creating and managing your account; authentication | Contract (Art. 6(1)(b)) |
| Projects & research data | Tool inputs and outputs you save while signed in | Cloud sync and collaboration | Contract (Art. 6(1)(b)) |
| AI feature inputs | Text, PDF content, or data you submit to AI features | Generating AI responses; processed transiently, not stored by us | Contract / Legitimate interest (Art. 6(1)(f)) |
| Feedback | Message, category, optional email and screenshot | Improving the platform; following up on bug reports | Legitimate interest (Art. 6(1)(f)) |
| Server logs | IP address, timestamp, user-agent, endpoint called | Security monitoring and operations; retained up to 90 days | Legitimate interest (Art. 6(1)(f)) |
3. Data We Do NOT Collect
- We do not store your own API key on our servers (it lives only in your browser)
- We do not use tracking cookies, advertising pixels, or third-party analytics beacons
- We do not sell, share, rent, or monetise user data in any form
- We do not retain AI input content on our servers after your request is fulfilled
- We do not collect race, ethnicity, health, biometric, or other special-category data
- We do not fingerprint browsers or build behavioural profiles for advertising
4. Local-first Architecture
All statistical calculations (meta-analysis, effect sizes, trend analysis, diagrams, NMA, RoB, GRADE, and every other computational tool) run entirely in your browser. No research data you enter into these tools is transmitted to our servers. Without an account, your projects are saved only in your browser’s local storage and never leave your device.
5. Third-Party Subprocessors
| Service | Purpose | Data sent | Location |
|---|---|---|---|
| Supabase | Account auth and cloud sync (account holders only) | Email, name, saved projects | Japan |
| OpenAI / Anthropic / Google | AI features (extraction, screening, advisor, assistant) | Your AI input text or PDF content (encrypted in transit, not retained by us) | United States |
| Google Cloud Run | Hosting the platform | Server logs only | United States |
| Lemon Squeezy / Stripe | Subscription payments | Billing details at checkout (we never receive your card number) | United States |
| NCBI E-utilities (PubMed) | Literature search | Your search query | United States |
| CDNs (Plotly, KaTeX, Lucide, SheetJS) | App assets and libraries | HTTP request headers only | Global CDN |
6. Data Retention
- Account and projects: Retained for as long as your account is active. You can delete your account and all associated data at any time.
- AI inputs: Kovex does not store AI inputs on its own servers after processing. AI inputs are transmitted to third-party AI providers to generate responses. Those providers may temporarily process or retain data according to their own policies and contractual terms.
- Server logs: Retained for up to 90 days for security and operational purposes, then deleted.
- Feedback submissions: Retained until resolved or no longer needed, and no longer than 2 years.
7. Sensitive Data Warning
Do not upload or save documents containing patient-identifiable information, personal health data, or any data subject to HIPAA, GDPR special-category protections, Taiwan PDPA sensitive categories, or equivalent regulations. Kovex is not designed or validated for handling sensitive personal data. Projects you save while signed in are stored in our cloud database, so keep them free of identifiable data and review your institutional data governance policies before use.
8. Your Rights under GDPR (EU / UK / EEA)
If you are located in the EU, EEA, or UK, you have the following rights under the General Data Protection Regulation (GDPR / UK GDPR):
- Right of access (Art. 15): You may request a copy of the personal data we hold about you.
- Right to rectification (Art. 16): You may ask us to correct inaccurate or incomplete personal data.
- Right to erasure (Art. 17): You may request deletion of your personal data. You can delete your account and all its data directly within the app at any time.
- Right to restriction of processing (Art. 18): You may ask us to restrict how we use your data in certain circumstances.
- Right to data portability (Art. 20): You may request your data in a structured, machine-readable format.
- Right to object (Art. 21): You may object to processing based on legitimate interests.
- Right to withdraw consent: Where processing is based on consent, you may withdraw it at any time without affecting the lawfulness of prior processing.
- Right to lodge a complaint: You have the right to lodge a complaint with your local supervisory authority (e.g., your national Data Protection Authority).
We respond to all GDPR data-subject requests within 30 days. To exercise any right, email help@covexe.com.
9. Your Rights under the Taiwan Personal Data Protection Act (PDPA)
If you are located in Taiwan, the Personal Data Protection Act (個人資料保護法, PDPA) applies. As the person in charge of personal data collection:
- Purpose of collection: Account and project management (purpose code: 040, 136 – Customer management; Internet service supply and information technology services).
- Categories of personal data: Identification data (name, email); online activity records (server logs); user-generated content (projects you save).
- Right to inquiry and review (Art. 10): You may request access to personal data we hold about you.
- Right to make copies (Art. 10): You may request copies of your personal data.
- Right to correction or supplement (Art. 11): You may request correction of incorrect data.
- Right to deletion (Art. 11): You may request deletion of your personal data where it is no longer required for the stated purpose or where you withdraw consent.
- Right to discontinue processing or use (Art. 11): You may request that we stop processing your data in certain circumstances.
We will respond to PDPA requests within 30 days. You may exercise these rights by contacting us at help@covexe.com. If you believe your rights have been violated, you may file a complaint with the Personal Data Protection Commission (PDPC) of Taiwan.
10. Your Rights under the CCPA (California Residents)
If you are a California resident, the California Consumer Privacy Act (CCPA) as amended by the California Privacy Rights Act (CPRA) grants you the following rights:
- Right to know: You may request disclosure of the categories and specific pieces of personal information we collect, the purposes for which we use it, and the categories of third parties with whom we share it.
- Right to delete: You may request deletion of personal information we have collected about you, subject to certain exceptions.
- Right to correct: You may request correction of inaccurate personal information.
- Right to opt out of sale or sharing: We do not sell or share personal information for cross-context behavioural advertising. You therefore need not opt out, but you may contact us to confirm this at any time.
- Right to limit use of sensitive personal information: We do not collect sensitive personal information as defined by the CPRA.
- Right to non-discrimination: We will not discriminate against you for exercising any CCPA rights.
California residents may submit a verifiable consumer request by emailing help@covexe.com. We respond within 45 days as required by law.
Categories of personal information collected in the past 12 months: Identifiers (name, email, IP address); Internet activity (server logs, pages visited); user-generated content (saved projects). No personal information has been sold or shared for advertising purposes.
11. International Data Transfers
Our platform is hosted in the United States. Account and project data is stored in Japan. AI features send your input to providers located in the United States. By using the Platform, you acknowledge that your personal data may be transferred to and processed in countries outside your own, including countries that may have different data protection standards. We rely on contractual and technical safeguards, including applicable transfer mechanisms, to protect cross-border data transfers where required.
12. Cookies and Tracking
Kovex does not use tracking cookies, advertising cookies, or third-party analytics scripts. We use only essential technical mechanisms (browser localStorage and sessionStorage) to keep your tool state and session active. These are not cookies and cannot be used for cross-site tracking.
13. Children’s Privacy
Kovex is not directed at children under the age of 16. We do not knowingly collect personal information from anyone under 16. If you believe a minor has provided us with personal data, please contact us immediately at help@covexe.com and we will delete it promptly.
14. Changes to This Policy
We may update this Privacy Policy from time to time. When we make material changes, we will update the “Last updated” date at the top and, where required by law, notify you by email or in-app notice. Continued use of the Platform after the effective date of any change constitutes your acceptance of the updated Policy.
15. Contact & Data Requests
For any privacy enquiry, data-subject request (access, correction, deletion, portability, objection), or complaint, contact us at:
Email: help@covexe.com
Subject line: “Privacy Request” or “Data Subject Request”
We will acknowledge your request within 72 hours and respond fully within the timeframe required by the applicable law (30 days for GDPR and Taiwan PDPA; 45 days for CCPA).
Data Processing Transparency
Kovex is committed to full transparency about how your data is handled at every step:
Open Methodology
Every statistical algorithm in Kovex is documented with its formula, source reference, and downloadable R replication code on the Methods page. This allows you to independently verify any result produced by Kovex using standard statistical software.
Intellectual Property & Citations
Third-Party Methodologies
Kovex implements the following published methodologies which are the intellectual property of their respective authors and organisations. Kovex is not affiliated with, endorsed by, or sponsored by any of these organisations:
- PRISMA 2020 - Page MJ et al., BMJ 2021;372:n71
- CONSORT 2010 - Schulz KF et al., BMJ 2010;340:c332
- RoB 2 - Sterne JAC et al., BMJ 2019;366:l4898
- ROBINS-I - Sterne JAC et al., BMJ 2016;355:i4919
- Newcastle-Ottawa Scale - Wells GA et al., Ottawa Hospital Research Institute
- QUADAS-2 - Whiting PF et al., Ann Intern Med 2011;155:529-536
- AXIS - Downes MJ et al., BMJ Open 2016;6:e011458
- GRADE - Guyatt G et al., J Clin Epidemiol 2011;64:383-394
- DerSimonian-Laird - DerSimonian R, Laird N. Control Clin Trials 1986;7:177-188
- Joinpoint Regression - Kim HJ et al., Stat Med 2000;19:335-351
Platform IP
The Kovex platform code, design, and user interface are proprietary. Unauthorised reproduction or distribution is prohibited.
Citing Kovex
When using Kovex in research, please cite the individual methodology tools used (see the All Citations page) and acknowledge Kovex as the platform. A formal platform citation will be available upon publication of the Kovex validation paper (in preparation).
Support
Get help with your account, report a problem, or find answers to common questions.
Figure Studio
Pick a pre-built figure type, map your columns, and get the same clean, publication-ready design every time. The layout, fonts, colours, and spacing are locked to a house style, so the figure is reproducible and consistent, not improvised.
What this does
Produces consistent, reproducible figures from pre-built templates. Unlike asking an AI to "draw a chart" (which differs every time), each template renders with a fixed design, so the same data always yields the same figure.
When to use
Whenever you need a clean standard figure (grouped bars with error bars, box plots, scatter with fit, multi-series line, ranked bars, histogram) without fiddling with styling.
How to use
- Paste data (first row = headers) or load the sample
- Choose a figure type
- Confirm the auto-mapped columns, or adjust them
- Set title and labels, pick a colour palette
- Export PNG or SVG
Interpreting results
Error bars show the column you map to "error" (for example a 95% CI half-width or SD). Scatter fit is ordinary least squares with the equation and R-squared shown.
More styling
Project Home
The home base for your active project. Every tool you use saves into this project automatically. Run blinded dual screening, resolve conflicts, and manage extraction templates here.
What this does
Your project workspace: a progress overview plus blinded dual screening, conflict resolution, and custom extraction templates. All tool data is auto-saved to the open project.
When to use
Throughout a review. Open a project, then work across the tools; come back here to screen references with two reviewers and resolve disagreements.
How to use
- Open or create a project (Projects button, top right)
- Use any tool and your inputs are saved to the project automatically
- Import references and have each reviewer vote independently
- Resolve conflicts and review inter-rater agreement (kappa)
- Build extraction templates and send them to the Extraction tool
Interpreting results
Cohen's kappa: <0.20 slight, 0.21-0.40 fair, 0.41-0.60 moderate, 0.61-0.80 substantial, >0.80 almost perfect agreement.
User Manual
A practical guide to KOVΞX LABS, organised by the systematic-review workflow. Each tool also has a "?" button on its own page for quick reference.
All Citations
Every methodology implemented in Kovex is backed by peer-reviewed literature. Copy individual citations or all at once for your Methods section.
What this does
Lists every academic reference backing the methods used in Kovex. Copy-ready for your manuscript.
When to use
When writing your methods or reference section and need to cite the statistical methods used.
How to use
- Browse citations by tool
- Click to copy individual citations
- Use "Copy All" for a complete reference list
Interpreting results
This is a reference page. Include relevant citations in your manuscript to document which methods and software were used.
A formal way to cite Kovex is on the way. We are preparing a methods paper that validates the platform's quantitative output against reference implementations; once it is published you will be able to cite it here.
My Plan
Your subscription, AI usage, and team. Use the preview switch to see what each plan looks like; once billing is live, your real plan drives this automatically.
PDF Library
Read your papers inside KOVEX, with no more juggling tabs. PDF, Word, text, Markdown and LaTeX are supported. Ask the paper questions, jump straight to extraction, and keep every reference for a project in one place. Files are stored privately in this browser.
API Key
AI features (data extraction, screening, the assistant, figures, and the data advisor) are free when you connect your own AI provider key. Your key is stored only in this browser and is sent with your AI requests. It is never saved on our servers.
PLANS
What this does
Shows available subscription plans and their included features.
When to use
When deciding which plan fits your research needs.
How to use
- Compare plan features
- Select the plan that matches your usage
Interpreting results
Free tier includes all core tools. Paid plans add AI-powered features and higher usage limits.
Is Kovex really free?
Yes. Every research tool is free for everyone, with no limits. AI features are free too when you connect your own API key.
What do I actually pay for?
Three optional things: managed AI (we run the AI so you do not need your own key), team collaboration with shared cloud projects, and institutional licenses.
Which AI providers can I use?
OpenAI, Anthropic (Claude), or Google (Gemini). Add your key on the AI Key page; it stays in your browser and is never stored on our servers.
Do you offer credits or academic deals?
Yes. Pay-as-you-go AI credits for occasional use, and discounted institutional licensing for universities. Contact us.