Where Should You Get Your Dental Thesis Analysed?
You have four realistic options for the statistics in a dental thesis: your university's biostatistics unit (often housed in the medical faculty), a faculty member with a statistics background, an individual freelance statistician, or an analysis service such as GetBayes. But in dentistry one technical question decides the choice: your observations are not independent. Several teeth from the same patient, two halves of the same mouth, repeated measurements on the same implant — an analysis that treats these as independent observations reports p-values that are smaller than they should be. Whoever you speak to, ask this first: "How will you handle multiple measurements from the same patient?"
Below are the trade-offs of each channel, the questions to ask about everything from split-mouth designs to examiner agreement, the warning signs, and price ranges compiled from public listings. GetBayes is one of these options and a paid service — we say plainly what we do at the end of the page.
Who is this guide for?
Residents writing a dental specialty thesis
PhD and master's students in dentistry
Clinicians running studies in implantology, periodontology, orthodontics, endodontics or restorative dentistry
Researchers running in-vitro studies (bond strength, microleakage, material comparison)
Anyone whose manuscript returned from a dental journal with statistical revisions
What makes dental data different
In most medical studies one patient is one row. In dentistry one patient can be 28 teeth, four implants, or two halves of a mouth. Measurements from the same mouth resemble each other (clustering), so 120 teeth collected from 30 patients are not 120 independent observations. An analysis that ignores this assumes a larger sample than you have, narrows confidence intervals and turns differences that are not there into significant findings. It is the single most common reviewer objection in dental manuscripts.
The right approach depends on the design: paired (dependent-samples) comparisons for split-mouth studies, approaches that model the clustering when many teeth or implants come from the same patient, and repeated-measures analysis for periodontal indices tracked over time. Whoever you hire, describe your data structure and hear their plan for this before work starts.
Where you can get the analysis done
| Channel | Strength | Weakness | Typical turnaround |
|---|---|---|---|
| University biostatistics unit | Methodologically the safest address; fluent in thesis and ethics requirements | Most dental faculties have no unit of their own and rely on the medical side; heavy workload means queueing | 2-8 weeks |
| Faculty member / senior resident with statistics background | Knows the clinical context and what the thesis requires | Carries their own workload; survival analysis and clustering can exceed the comfort zone | Variable, often open-ended |
| Individual / freelance statistician | Flexible, negotiable | May have no experience with split-mouth or tooth-level data — this is where the classic mistake of treating every tooth as independent appears | 3 days - 2 weeks |
| Professional analysis service (like GetBayes) | Written fixed quote, journal-format reporting, free revisions and post-delivery Q&A | Paid; with a zero budget and plenty of time, learning it yourself may fit better | Analysis in 15 minutes, same-day delivery |
Seven questions to ask before you commit
The answers predict the reviewer comments you will receive:
How will you handle multiple teeth or implants from the same patient? You need to hear that clustering will be accounted for; "we count each tooth as a separate observation" is the most common and most costly error.
Which test will you use for a split-mouth design? Because both halves belong to the same patient, the comparison must be paired: dependent-samples t-test or Wilcoxon signed-rank, not an independent-groups test.
How will implant or restoration success be reported? With variable follow-up, a percentage success rate is not enough; Kaplan-Meier survival curves, log-rank tests and Cox regression for risk factors are expected.
Which coefficient will you use for examiner agreement? Cohen's kappa for categorical assessments (weighted kappa for ordinal), the intraclass correlation coefficient (ICC) for continuous measurements, and a Bland-Altman plot where relevant. Manuscripts submitted without calibration data come back with revisions.
How will count data such as DMFT be analysed? Caries counts are skewed and zero-inflated; they call for nonparametric tests or count models, and a t-test is often not appropriate.
Will multiple comparisons be corrected in the in-vitro study? Comparing four or five material groups pairwise without correction manufactures false positives; they should be able to name the post-hoc method.
Was the sample size calculated at tooth level or patient level? If this is not settled at the protocol and ethics stage, it cannot be fixed after data collection.
Red flags
"Let's use the number of teeth as n so the sample looks bigger." Statistically indefensible, and reviewers ask about it in the first round.
A promise that results will come out significant. Significance cannot be promised. A null result is a valid finding and, written properly, passes a defence.
Being told examiner agreement "isn't needed." Nearly every study involving clinical indices or radiographic measurement is expected to report it.
A quote given without asking about the study design. A split-mouth study and a parallel-group study are not the same analysis.
Raw SPSS output with the interpretation left to you. What you present at your defence is the interpretation, not the output.
No written confidentiality commitment. De-identify patient data before sending; names, IDs and file numbers are not needed for analysis.
What does it cost? (2026 market ranges)
Compiled from public service listings; these are not GetBayes prices.
| Scope of analysis | Market range | Note |
|---|---|---|
| Descriptive statistics + group comparisons | $300 - $600 per project | Narrow-scope in-vitro studies and simple clinical comparisons |
| Repeated measures, examiner agreement, multivariable analysis | $600 - $1,000 per project | Where a typical dental specialty thesis lands |
| Survival analysis, clustered data modelling, multicentre studies | $1,000 - $1,500+ per project | Implant follow-up studies and complex designs |
| Reviewer revisions / additional analyses | Usually billed separately across the market | At GetBayes, revisions driven by a reviewer, advisor or committee are free |
What to send in your first message
- 01
Unit of observation
What does one row of your table represent — a patient, a tooth, an implant, a time point? This single fact determines half of the analysis plan.
- 02
Design
Split-mouth, parallel-group or in-vitro; how many groups, and at how many time points measurements were taken.
- 03
De-identified data
Your Excel or SPSS file with patient names and file numbers removed. Missing values and messy coding are fine.
- 04
Timeline and target format
Your submission or defence date and, if relevant, the journal you plan to submit to — tables are built accordingly.
The ethical line, and where we stand
Analysis support is common and accepted in dental research, and the boundary is clear: the data must be real and yours, and you must understand and be able to defend the methods. A good report therefore justifies every methodological choice alongside the result — the answer to "why this test?" should already be in your report.
GetBayes is one of the options above: we run comparisons that correctly model dependency in split-mouth and tooth-level data, Kaplan-Meier and Cox for implant and restoration survival, kappa/ICC for examiner agreement, repeated-measures analysis for periodontal indices, and appropriate models for count data such as DMFT. The analysis itself usually takes 15 minutes; delivery is same-day, often within hours or even minutes. Revisions driven by an advisor, committee or reviewer are free.
Frequently asked questions
My sample is 30 patients but 120 teeth — which is n?
Report both, but build the statistics around the actual unit of analysis. Teeth from the same mouth resemble each other, so 120 teeth are not 120 independent observations; ignoring clustering produces p-values that are too small. State "30 patients, 120 teeth" explicitly in the table and model the dependency in the analysis.
Which test should a split-mouth study use?
Because both halves come from the same patient, the comparison must be paired: a dependent-samples t-test when normality holds, the Wilcoxon signed-rank test when it does not. Using an independent-groups test throws away the design's greatest strength — within-patient control.
Why is examiner agreement required, and how is it calculated?
Clinical indices and radiographic measurements vary between examiners, so journals want evidence that the measurement is reproducible. Cohen's kappa is used for categorical assessments (weighted kappa for ordinal scales) and the intraclass correlation coefficient (ICC) for continuous ones, with a Bland-Altman plot to show systematic bias where needed. Planning this at the protocol stage is far easier than re-measuring afterwards.
My in-vitro study compares four materials — what analysis do I need?
One-way ANOVA with an appropriate post-hoc test (Tukey, for example) when assumptions hold; Kruskal-Wallis with corrected pairwise comparisons when they do not. The critical point is multiple-comparison correction: comparing four groups pairwise without it manufactures false positives. If you also split groups into strata (before/after ageing, say), a two-way ANOVA may fit better.
Is it safe to share patient data?
Share de-identified data; names, IDs and file numbers are not needed for analysis. At GetBayes your data is used only for your study, never shared with third parties, and automatically deleted no later than 90 days after final delivery — you can always request earlier deletion.
How long does it take, and how is the price set?
The analysis itself usually takes 15 minutes and delivery is same-day. The price is set by the complexity of the design and the number of analyses; it does not change with delivery speed or the number of revisions. Send your data and you get a free assessment and a clear written quote within 24 hours.
Get a free assessment for your dental thesis
Send your unit of observation, your design and your de-identified data — we'll reply within 24 hours with a free assessment.
Last updated: August 31, 2026