Experimental_Design_in_Biomedical_Sciences_6-2024

Experimental Design in Biomedical Sciences

Measurements Overview

  • Presenter: Eva Szegezdi

  • Importance of measurements in experimental design

Taking Measurements

  • Precise Measurements:

    • Reduce variation and increase accuracy.

  • Replicates Example:

    • Data from various mg/ml concentrations:

      • 1 mg/ml: 0.140, 0.153, 0.166

      • 2 mg/ml: 0.350, 0.382, 0.416

      • 4 mg/ml: 0.720, 0.785, 0.855

    • Identifying imprecision versus inaccuracy.

Imprecision

  • Definition:

    • Errors that are uncorrelated between measurements.

    • Each measurement can be either over or underestimated.

  • Issue:

    • Reproducibility can be problematic.

  • Example:

    • Measuring walking time with analogue vs digital stopwatch.

Inaccuracy and Bias

  • Definition:

    • Errors that are correlated across measurements.

    • Systematic error pattern observed.

  • Example:

    • Using a consistently fast-running watch.

  • Importance of calibration checks.

Bias in Data Collection

  • Bias Definition:

    • Deviation from truth leading to false conclusions.

  • Types:

    • Intra-observer variability: Human error leading to inconsistent data.

  • Ways to Avoid Bias:

    • Introduce objective descriptions for categories.

    • Maintain a portfolio of example cases near category borders.

    • Conduct repeatability studies.

Example of Intra-observer Variability

  • Scoring System:

    • 0 = no facial hair

    • 1 = small moustache

    • 2 = big moustache

    • 3 = moustache with small beard

    • 4 = moustache with big beard

Repeatability Study

  • Definition:

    • Procedure to measure random samples multiple times and compare values.

  • Goal:

    • Achieve consistency in repeated measurements.

  • High repeatability indicates low imprecision, but doesn’t guarantee the absence of bias.

  • Checking Bias:

    • Use samples with known true measurements for blind scoring.

    • Collaborate with another observer to compare results.

Precision vs. Accuracy

  • Graphical Representation:

    • Low accuracy vs. high accuracy.

    • Systematic errors illustrated across different scenarios.

Inter-observer Variability

  • Definition:

    • Variability arising from different observers scoring the same sample differently.

  • Minimization Techniques:

    • Directly compare results among observers.

    • Establish clear measurement rules and methods.

    • Normalize data when applicable.

Observer Effects

  • Observation Impact:

    • The act of observing can influence the behavior of biological systems.

  • Collection Considerations:

    • Explore alternatives like written questionnaires or video recordings.

  • Pilot Studies:

    • Test data collection methods before full implementation.

Floor and Ceiling Effects

  • Floor Effect:

    • Majority of measurements at the lowest range.

  • Ceiling Effect:

    • Majority of measurements at the highest range.

Sensitivity, Specificity, and Accuracy

  • Sensitivity:

    • Ability to detect true positives; vital statistics for diagnostic tests.

      • Example: 85% sensitivity means 85% of affected individuals test positive.

  • Specificity:

    • Proportion of true negatives correctly identified; e.g., 95% specificity means 95% of unaffected individuals test negative.

  • Accuracy Calculation:

    • The ability to identify samples correctly:

    • Accuracy = (TN + TP)/(TN + TP + FN + FP)

Calculating Sensitivity and Specificity

  • Formulas:

    • Sensitivity = TP / (TP + FN)

    • Specificity = TN / (TN + FP)

Today's Task

  • Task 1:

    • Open file "BM3101_Week 7_Task1.xlsx".

    • Repeat calculations of sensitivity and specificity from "Example_data_sensitivity_specificity.xlsx".