Statistician for dental researchers: inter-rater agreement (kappa and ICC), logistic regression, survival analysis and reviewer responses. PhD theses and Q1/Q2.
Statistics for dental research that understands teeth nested in patients
PhD theses, Q1/Q2 articles and reviewer responses in dentistry and oral health. ICC and kappa for inter-rater calibration, mixed models for nested data (teeth within patients), in vitro studies, small clinical trials and meta-analysis.
Dentistry has very specific statistical challenges. If yours fits any of these, we are a fit.
You need to report kappa or ICC between raters before the main study. You get 0.49 (fair) and the journal requires ≥0.75. Recalibrate? Different model? There's method.
You analyzed each tooth as independent and the reviewer reminds you they aren't: 28 teeth belong to the same patient. You need mixed models with patient as random effect.
You tested 3 materials × 10 specimens and need correct multiple comparisons, ANOVA with appropriate corrections, non-parametric alternatives if needed and effect sizes.
Correct ICC model (1,1 vs 2,1 vs 3,1) per design, weighted kappa for ordinal scales, Fleiss kappa for >2 raters and clinically useful SEM.
Teeth within patients, sites within mouth, lesions within patient. Random effects, patient-level ICC and correct reporting.
ANOVA with post-hoc (Tukey, Bonferroni), Kruskal-Wallis with Dunn, equivalence and non-inferiority if your hypothesis demands it.
Comparing two diagnostic methods (radiograph vs CBCT, two measurement software): mean bias, limits of agreement and proper Bland-Altman plot.
Kaplan-Meier for implant survival, Cox for failure risk factors and correct follow-up reporting.
Forest plots, I² heterogeneity, random-effects models for heterogeneous samples (in vitro vs clinical) and publication bias.
No clock running on your tab. We agree on scope, timeline and a fixed price upfront: no inflated hours, no surprise invoice at the end. Before delivery, every analysis goes through a careful review on my side: I rerun the code, re-check the assumptions and review tables and figures line by line. That's the part I obsess over: that what you receive is ready to defend without errors surfacing in peer review. Well-bounded projects (inter-rater calibration, three-material in vitro comparison, reviewer response) fit this model. If something new comes up, we discuss it before I touch it.
Illustrative example: survival curves (e.g. implant or restoration survival) with log-rank test and hazard ratio. Useful for follow-up data with censoring, common in dental research.
Free resources on the analyses most requested in dental research.
Inter-rater calibration, nested data, in vitro, reviewer response, meta-analysis… Fill the form and I respond within 24h with a tentative proposal.
Biostatistics and statistical consulting for health research · All services
Before discarding the study: we check whether the kappa model is correct (sometimes simple kappa is used when weighted kappa or ICC is needed), whether the scale can be collapsed into broader categories, and whether one extra calibration session changes things. I've seen 0.55 → 0.82 just by adjusting the model and one extra session.
If your clinical unit is the tooth (caries, restoration, acid attack), analyzing each tooth as independent overestimates power and the reviewer will flag it. The right solution is mixed models with patient as random effect. I deliver both analyses so you see the difference and justify it.
Depends on variable type and assumptions. Continuous with normality and homoscedasticity: ANOVA + Tukey. Without normality: Kruskal-Wallis + Dunn. N=10 per group is tight but defensible if the report includes effect size and explicit assumptions.
Yes, it's one of the most requested services. I review the reviewer letter, run the additional analyses and help you draft the point-by-point response. Turnaround: 48-72h.