Biological Data Analysis

Vakinformatie

Nieuw vak vanaf 2025-2026.

Examenvragen

16/12/2025

PART 1: PRACTICAL EXERCISE IN R

Biological scenario

Researchers are interested in whether artificial light at night affects circulating melatonin levels in nestling songbirds. Twelve nest boxes of great tits with clutches of similar age were selected in a forest. At day 10 post-hatching, each nest box was randomly assigned to one of four night-time light regimes: control (no artificial light), low light, medium light or high light intensity. For the next 7 nights, LED lights were switched on at sunset and off at sunrise according to the assigned regime. At day 17 post-hatching, eight randomly chosen nestlings per nest were bled, and plasma melatonin concentration (in pg/mL) was quantified using a radio-immunoassay. You are given a dataset (melatonin_light_pollution.csv) where each line is one chick and the following variables:

• NEST – nest identity (factor with 12 levels)

• TREATMENT – factor with four levels: "control", "low", "medium", "high"

• MELATONIN – plasma melatonin concentration (continuous, pg/mL).

The a priori biological hypothesis is that any artificial light treatment reduces melatonin levels relative to the dark control.

Question

Test statistically which light regimes significantly reduce melatonin levels compared to the control treatment, using the most appropriate statistical model for this dataset.

In your answer:

• Motivate your choice of model type (including how you deal with nest identity).

• Fit the model and provide the relevant summary tables.

• Carry out appropriate posthoc tests to compare each light treatment to the control while correcting for multiple comparisons.

• Test if residuals are normally distributed and if there are any outliers or influential observations.

• Provide a graphical representation of the fitted effects.

• Interpret your findings in biological terms and provide a final analysis conclusion as you would report it in a scientific article.

Include your R code and the relevant numerical and graphical output in your submission (all pasted in this Word document).

PART 2: THEORY / INSIGHT QUESTIONS

1. A researcher studies the effect of temperature on the metabolic rate of lizards. He has two large incubators. He sets Incubator A to 20°C and Incubator B to 30°C. He places 20 lizards in Incubator A and 20 lizards in Incubator B. After one hour, he measures the metabolic rate of all 40 lizards. He analyzes the data using an unpaired t-test comparing the 20 values from A vs. the 20 values from B and finds a significant difference (p< 0.001). He concludes that temperature definitely caused the difference. What is the major flaw in this experimental design and analysis?

2. Immunologists test the efficacy of three antiviral drugs (A, B, C) against an untreated control in cell culture. For each condition (control, A, B, C) they plate infected cells into 8 culture wells. After 24 hours they examine 100 cells per well and record how many are still infected (INFECTED). They calculate the proportion infected per well (INFECTED/100) and fit a linear model with this proportion as the response and DRUG as a factor (4 levels), assuming normal errors. Tukey posthoc tests are used to compare all pairs of treatments. Discuss shortcomings of this approach and suggest a more appropriate statistical model. Mention both the error distribution and how the data should be entered.

3. A study calculates the correlation between 50 different environmental variables (pH, humidity, soil nitrogen, etc.) and the abundance of a specific rare mushroom. The authors perform 50 separate Pearson correlation test. They find that "Soil calcium" and "Wind speed" are significantly correlated with abundance (p= 0.04 and p=0.03) and publish a paper focusing on these two drivers. They do not mention the other 48 tests. Identify the statistical fallacy and the appropriate remedy.