Week 3 | Identify variables & Tidy data

Class Details

📅 Date: 20 March, 2025
⏰ Time: 09:30h - 11:30h
📖 Synopsis: Extract experimental details from text, format tidy tables, and reverse engineer study design from tidy tables.


Experimental Designs

1. Controlled Experiment

  • Description: Researcher manipulates one or more variables while keeping others constant.
  • Example: Adjusting pH levels in aquaria and measuring coral growth.
  • Strengths: Clear cause-effect relationships, high internal validity.
  • Limitations: May lack realism in natural environments.

2. Observational Study

  • Description: No manipulation; variables are observed as they naturally occur.
  • Example: Comparing fish diversity in protected vs. unprotected areas.
  • Types:
    • Cross-sectional: Single point in time.
    • Longitudinal (Time-series): Data collected over time.
  • Strengths: Realistic, practical.
  • Limitations: Cannot infer causation definitively.

3. Correlational Study

  • Description: Measures the relationship between variables without manipulation.
  • Example: Monitoring algal biomass and oxygen levels to detect patterns.
  • Strengths: Good for detecting associations.
  • Limitations: No causal conclusions.

4. Randomized Controlled Trial (RCT)

  • Description: Random assignment of subjects/units to treatments and control.
  • Example: Randomly assigning noise levels to dolphin enclosures.
  • Strengths: Minimizes bias, strong causal inference.
  • Limitations: Logistically complex in the field.

5. Factorial Design

  • Description: Tests multiple independent variables simultaneously.
  • Example: Testing combined effects of temperature and salinity on zooplankton.
  • Strengths: Identifies interaction effects between variables.
  • Limitations: Requires more samples.

6. Before-After Control-Impact (BACI) Design

  • Description: Compares conditions before and after an impact at both affected and control sites.
  • Example: Measuring water quality before and after coastal development.
  • Strengths: Controls for natural variability.
  • Limitations: Requires baseline data.

7. Natural Experiment

  • Description: Utilizes naturally occurring events as treatments.
  • Example: Studying fish populations after a hurricane.
  • Strengths: High ecological validity.
  • Limitations: Limited control over variables.

Goal | This exercise aims at training students to critically extract key experimental details from textual descriptions.

Instructions | Read each narrative description carefully. For each one, identify:

  • Experimental Design (check the Experimental Designs tab)
  • Independent Variable(s)
  • Dependent Variable(s)
  • Control Variable(s) (if applicable)

Example 1: Ocean Acidification on Coral Growth

Researchers want to test how increased ocean acidification affects the growth of a common coral species. They set up three aquaria with seawater at different pH levels: 8.1 (current ocean pH), 7.8, and 7.5. Identical coral fragments are placed in each aquarium, and all tanks are kept under the same light and temperature conditions. After six months, they measure the calcification rate of the corals.

  • Experimental Design: Controlled laboratory experiment with three treatment groups.
  • Independent Variable: pH level of seawater (8.1, 7.8, 7.5)
  • Dependent Variable: Coral calcification rate (growth)
  • Control Variables: Light intensity, water temperature, coral species and fragment size, nutrient levels.

Example 2: Marine Protected Areas and Fish Biodiversity

A team surveys fish species diversity in 10 marine protected areas (MPAs) and compares it with 10 nearby unprotected coastal sites. Surveys are conducted using the same underwater visual census method at all locations during the same season.

  • Experimental Design: Observational comparative study (between protected and unprotected areas)
  • Independent Variable: Protection status (MPA vs. unprotected site)
  • Dependent Variable: Fish species diversity (number of species observed)
  • Control Variables: Survey method, season, location depth, time of day surveyed.

Example 3: Plastic Pollution Impact on Zooplankton

Scientists investigate whether microplastic concentration affects zooplankton reproduction. They prepare four tanks with different microplastic concentrations: 0 particles/mL, 10 particles/mL, 100 particles/mL, and 1,000 particles/mL. Each tank contains the same number of zooplankton individuals and is kept at constant temperature, salinity, and food availability. After two weeks, they count the number of offspring produced.

  • Experimental Design: Controlled laboratory experiment with graded treatments.
  • Independent Variable: Microplastic concentration (0, 10, 100, 1,000 particles/mL)
  • Dependent Variable: Number of zooplankton offspring
  • Control Variables: Temperature, salinity, food availability, initial zooplankton number/species.

Example 4: Algal Bloom Effects on Oxygen Levels

To examine the effect of algal blooms on oxygen levels, researchers monitor dissolved oxygen in a coastal bay over six months. They record algal biomass, water temperature, and nutrient concentrations weekly. They look for correlations between algal biomass and oxygen concentration.

  • Experimental Design: Observational time-series study
  • Independent Variable: Algal biomass (measured continuously)
  • Dependent Variable: Dissolved oxygen concentration
  • Control Variables: Recording temperature and nutrient levels allows adjustment for confounding variables.

Example 5: Noise Pollution Impact on Dolphin Communication

Marine biologists play varying levels of ship noise recordings in a controlled enclosure housing a pod of dolphins. For each noise level (low, medium, high), they observe and record the frequency and duration of dolphin whistles. Sessions are randomized, and the dolphins have rest periods between exposures.

  • Experimental Design: Randomized controlled experiment
  • Independent Variable: Ship noise level (low, medium, high)
  • Dependent Variables: Frequency and duration of dolphin whistles
  • Control Variables: Same dolphin pod, rest periods, enclosure size, time of day.

Goal | To bridge experimental design concepts with practical data skills.

Instructions | After identifying the independent, dependent, and control variables from each “narrative”, describe how the data would be organized in a tidy data table. Specifically:

  1. Columns: What variables are measured or recorded?
  2. Rows: What does each row represent (unit of observation)?
  3. Variable types: Are variables numeric or categorical?
  4. Control variables: How are they documented?

Example 1: Ocean Acidification on Coral Growth

Tidy Table Example:

tank_id pH calcification_rate light_intensity temp_C coral_species nutrient_conc
Tank_1 8.1 2.3 200 25 Acropora sp. 3
Tank_2 7.8 1.9 200 25 Acropora sp. 3
Tank_3 7.5 1.4 200 25 Acropora sp. 3

Each row = one coral fragment after six months.

Example 2: Marine Protected Areas and Fish Biodiversity

Tidy Table Example:

site_id protection_status fish_species_count survey_method season depth_m time_of_day
MPA_1 MPA 25 Visual Census Spring 10 Morning
MPA_2 MPA 30 Visual Census Spring 12 Morning
Site_1 Unprotected 18 Visual Census Spring 11 Morning
Site_2 Unprotected 20 Visual Census Spring 10 Morning

Each row = one survey at one site.

Example 3: Plastic Pollution Impact on Zooplankton

Tidy Table Example:

tank_id microplastic_conc offspring_count temp_C salinity food_available zooplankton_sp initial_count
Tank_A 0 45 22 35 High Copepod sp. 50
Tank_B 10 40 22 35 High Copepod sp. 50
Tank_C 100 28 22 35 High Copepod sp. 50
Tank_D 1000 12 22 35 High Copepod sp. 50

Each row = one tank’s measurement.

Example 4: Algal Bloom Effects on Oxygen Levels

Tidy Table Example:

date algal_biomass dissolved_oxygen water_temp_C nutrient_conc
2024-01-01 12.5 6.8 18.2 3.1
2024-01-08 15.3 5.4 18.4 3.2
2024-01-15 20.1 4.7 18.6 3.0

Control variables (temperature and nutrients) are recorded alongside the key variables to control for confounding effects.

Example 5: Noise Pollution Impact on Dolphin Communication

Tidy Table Example:

session_id noise_level whistle_freq whistle_duration rest_period_min pod_id enclosure_size_m time_of_day
Sess_1 Low 7500 2.1 30 Pod_1 50 Morning
Sess_2 Medium 7100 1.8 30 Pod_1 50 Morning
Sess_3 High 6800 1.4 30 Pod_1 50 Morning

Each row = one noise exposure session.

Goal | To “reverse engineer” the experimental design and research question from a tidy table to build critical thinking skills.

Instructions | In this exercise, you are given tidy data tables. Your task:

  1. Interpret: What scientific question might the researchers be asking?
  2. Identify Variables: Which variables are:
    • Independent (manipulated or compared)
    • Dependent (measured outcome)
    • Control (kept constant or recorded for adjustment)
  3. Describe Experimental Design: Is it observational, controlled, randomized?

Example 1

tank_id light_intensity nutrient_level temp_C seaweed_species biomass_increase
Tank_1 100 Low 18 Ulva sp. 4.5
Tank_2 100 High 18 Ulva sp. 8.1
Tank_3 200 Low 18 Ulva sp. 6.2
Tank_4 200 High 18 Ulva sp. 11.4

Possible Research Question:

How do light intensity and nutrient levels affect seaweed biomass increase under controlled conditions?

Experimental Design:
Controlled laboratory factorial experiment (2x2 design: light × nutrient).

Variables:

  • Independent Variables: light_intensity (100, 200), nutrient_level (Low, High)
  • Dependent Variable: biomass_increase
  • Control Variables: temperature, seaweed_species (constant across tanks)

Example 2

site_id fishing_pressure depth_m habitat_type survey_method shark_count
Site_A High 25 Reef Baited Camera 3
Site_B Low 25 Reef Baited Camera 8
Site_C High 30 Sandy Baited Camera 2
Site_D Low 30 Sandy Baited Camera 6

Possible Research Question:

Is shark abundance lower in areas with high fishing pressure compared to low fishing pressure areas?

Experimental Design:
Observational comparative study (across sites).

Variables:

  • Independent Variable: fishing_pressure (High, Low)
  • Dependent Variable: shark_count
  • Control Variables: depth_m, habitat_type, survey_method (standardized across sites)

Example 3

sample_id salinity temp_C sample_depth_m chlorophyll_conc plankton_sp_count
Sample_1 30 22 5 2.5 15
Sample_2 35 22 5 3.1 18
Sample_3 40 22 5 4.2 12
Sample_4 45 22 5 4.8 9

Possible Research Question:

How does salinity influence plankton species diversity in coastal waters?

Experimental Design:
Observational gradient study.

Variables:

  • Independent Variable: salinity (continuous gradient)
  • Dependent Variable: plankton_species_count
  • Control Variables: temperature, sample_depth_m, chlorophyll_concentration (recorded)

1. Experimental Design & Variables

  • What information helps you distinguish between independent, dependent, and control variables?
  • How can you tell whether a study is controlled, randomized, or observational based solely on the dataset?
  • Why is it important to clearly document control variables, even if they are kept constant?
  • In some examples, more than one independent variable is manipulated. How would you recognize and interpret this factorial design?
  • How might confounding variables influence the results if not properly controlled?

2. Tidy Data Structure

  • Why is it useful to structure experimental data in a tidy format (one variable per column, one observation per row)?
  • What could go wrong in analysis or visualization if data were not kept in tidy format?
  • How do identifiers like tank_id, site_id, or session_id help maintain data clarity and traceability?

3. Critical Interpretation

  • Could there be alternative research questions addressed with the same dataset? If so, how?
  • Are there any additional variables you might want to collect to strengthen the experimental design or account for potential confounders?
  • If you encountered missing or inconsistent data in these tables, how would it affect your interpretation?
  • How do replication and sample size considerations appear (or not appear) in tidy datasets like these?
  • How could these tables be prepared for statistical modeling (e.g., regression analysis)?

4. Future Application

  • In your own research, how would you apply what you’ve learned about tidy data and variable types to plan data collection?
  • What challenges might arise when moving from narrative descriptions of experiments to data analysis-ready datasets in real-world projects?
Back to top