A medical specialty thesis is a resident's first major independent study in their academic career. For the resident physician left alone with the statistics section amid the intensity of clinical rotations, this guide aims to clarify the entire process.
Before You Begin: Is the Study Design Correct?
The vast majority of statistical problems emerge after the data has been collected; yet the solution lies in the design decisions that must be made before data collection.
The most common designs in specialty theses are as follows:
Retrospective cohort: Data is collected retrospectively from hospital archives. The ethics committee process is shorter and data collection is faster. However, missing data, inconsistent records, and confounding factors are frequently encountered problems.
Prospective observational: Patients are followed prospectively. It is a stronger design but requires time; whether it can be completed within the residency period must be carefully calculated.
Case-control: Useful for rare diseases. An odds ratio is calculated, but a causal relationship cannot be established.
Randomized controlled trial: It is the gold standard but is rarely preferred for a specialty thesis; it requires ethics approval, randomization, and an adequate follow-up period.
Sample Size: The Most Frequently Skipped Step
This question, which is always asked on the ethics committee form, is the part residents struggle with the most.
Three parameters are needed for a power analysis:
- Effect size: How large is the expected difference? It is determined from similar studies in the literature.
- Alpha (α): The probability of a Type I error, usually 0.05
- Power (1-β): Usually 80% or 90%
G*Power is a free and widely used tool. Selecting the correct calculation method according to your study design, whether an independent two-group t-test, survival analysis, or logistic regression, is critically important.
Which Analyses Are Performed?
The statistical methods most commonly used in specialty theses:
Descriptive statistics: Mean, standard deviation, median, IQR, percentage distributions. Present in every study, with tables presented in APA or Vancouver format.
Normality test: Shapiro-Wilk (for small samples) or Kolmogorov-Smirnov. Determines whether a parametric or non-parametric test will be used.
Two-group comparison: Independent samples t-test (if the distribution is normal) or Mann-Whitney U (non-parametric).
More than two groups: ANOVA or Kruskal-Wallis, followed by post-hoc tests.
Categorical variables: Chi-square or Fisher's exact test.
Correlation: Pearson (parametric) or Spearman (non-parametric).
Survival analysis: Kaplan-Meier and the log-rank test, Cox regression.
Multivariable analysis: Logistic regression or Cox regression; to control for confounding factors.
The Questions the Jury Asks Most
The questions most frequently asked about the statistics section during specialty thesis defenses are as follows:
"How did you calculate the sample size?"; You should be able to explain the power analysis.
"Why did you choose this test?"; If a normality test was performed, you should be able to show that the decision was made accordingly.
"How did you control for confounding factors?"; You should be able to explain why the multivariable analysis was performed.
"Why did you choose X for the post-hoc test?"; You need to know the difference between Bonferroni, Tukey, and LSD.
Tables and Figures
For medical specialty theses, tables are usually placed within the thesis as a Word document. The resolution of figures should be at least 300 DPI.
Visuals such as box plots, Kaplan-Meier curves, or ROC curves can be easily created with R or SPSS. Knowing which visual to use where directly affects the presentation quality of the thesis.
If you want to complete your specialty thesis statistics process faster and more accurately, you can request a free 30-minute assessment consultation.
Where Do People Get Stuck Most in This Analysis?
- You need to perform a power analysis but do not know where to find the effect size.
- The jury will ask "why did you choose this test," but which path to take after the normality test is unclear.
- The tables are ready but you do not know how to write the interpretation section.