Descriptive Statistics, Hypothesis Testing, and Inferential Analyses in Biology

Introduction to Biological Data and Central Tendency

• Population Distributions and Sample Size:

  • Biological data ideally forms a symmetrical bell-shaped curve (normal distribution) when representing a whole population.

  • Small sample sizes tend to produce jagged, uneven frequency distributions with high variability.

  • Increasing sample sizes smooths out data fluctuations, yielding a representative bell curve.

• Sample Data Structure (Walleye Length Example):

  • Variable measured: Walleye (Sander vitreus) length.

  • Unit of measurement: Centimeters ($cm$).

  • Data organization: Structured in tabular format (rows and columns) across two distinct locations:

    • Oneida Lake

    • Delta Lake

• Role of Descriptive Statistics:

  • Descriptive statistics summarize and highlight key patterns within a dataset.

  • Primary analytical objectives:

    1. Determine the center of the dataset (central tendency / peak of the bell curve).

    2. Quantify the dispersion or spread of the data around that center.

• Measures of Central Tendency:

  • Mode:

    • Definition: The most frequently occurring observation in a dataset.

    • Represents the highest peak of a frequency distribution curve.

    • Less frequently emphasized in basic biological descriptive statistics.

  • Median:

    • Definition: The exact middle value in an ordered dataset.

    • Represents the $50\text{th}$ percentile point where $50\%$ of observations lie below and $50\%$ lie above.

  • Mean (Sample Average, $\bar{x}$):

    • Definition: The arithmetic average of all observations in a sample.

    • Symbol: Variable $x$ with an overbar ($\bar{x}$).

    • Mathematical Formula:             $\bar{x} = \frac{\sum_{i=1}^{n} x_i}{n}$             where $\sum x_i$ is the sum of all individual observations and $n$ is the total sample size.

• Calculation of Means for Walleye Data:

  • Oneida Lake Dataset:

    • Sum of length measurements ($\sum x$): $774\,cm$

    • Sample size ($n$): $10$

    • Sample Mean Calculation:             $\bar{x}_{\text{Oneida}} = \frac{774\,cm}{10} = 77.4\,cm$

  • Delta Lake Dataset:

    • Sum of length measurements ($\sum x$): $704\,cm$

    • Sample size ($n$): $10$

    • Sample Mean Calculation:             $\bar{x}_{\text{Delta}} = \frac{704\,cm}{10} = 70.4\,cm$

  • Comparative Conclusion: On average, walleye sampled from Oneida Lake ($77.4\,cm$) are longer than those sampled from Delta Lake ($70.4\,cm$).

Measures of Data Dispersion: Variance and Standard Deviation

• Importance of Quantifying Spread:

  • Two distinct datasets can share an identical measure of central tendency (e.g., $\bar{x} = 50$) while possessing entirely different shapes and spreads:

    • Narrow, steep curve with minimal variation around the mean.

    • Moderate curve with moderate dispersion.

    • Broad, flat curve with high dispersion spread widely across the scale.

  • Evaluating central tendency alone is insufficient to describe a biological population.

• Sample Variance ($s^2$):

  • Definition: The average squared deviation of each observation from the sample mean.

  • Purpose: Measures the absolute spread of data points relative to the mean center.

  • Units: Units are squared relative to original measurements (e.g., $cm^2$).

  • Mathematical Formula:         $s^2 = \frac{\sum_{i=1}^{n} (x_i - \bar{x})^2}{n - 1}$

  • Component breakdown:

    • $x_1, x_2, \dots, x_n$: Individual observations.

    • $\bar{x}$: Calculated sample mean.

    • $n - 1$: Degrees of freedom (for $n = 10$, denominator is $10 - 1 = 9$).

    • Expanded numerator calculation: $(x_1 - \bar{x})^2 + (x_2 - \bar{x})^2 + \dots + (x_n - \bar{x})^2$.

• Sample Standard Deviation ($s$):

  • Definition: The square root of the sample variance.

  • Purpose: Converts the measure of dispersion back into original measurement units (e.g., $cm$ instead of $cm^2$) for intuitive interpretation.

  • Mathematical Formula:         $s = \sqrt{s^2} = \sqrt{\frac{\sum_{i=1}^{n} (x_i - \bar{x})^2}{n - 1}}$

• Calculations for Walleye Standard Deviation:

  • Oneida Lake:

    • Sum of squared deviations ($\sum (x_i - \bar{x})^2$): $398.4$

    • Sample Variance ($s^2$):             $s^2 = \frac{398.4}{10 - 1} = \frac{398.4}{9} \approx 44.3\,cm^2$

    • Sample Standard Deviation ($s$):             $s = \sqrt{44.3\,cm^2} \approx 6.65\,cm$

    • Rounding: Depending on significant figure rules, $6.65\,cm$ can be rounded to $7\,cm$ (using 1 significant digit) or kept at $6.65\,cm$ (using 2 decimal places).

  • Delta Lake:

    • Sample Variance ($s^2$): $40.9\,cm^2$

    • Sample Standard Deviation ($s$):             $s = \sqrt{40.9\,cm^2} \approx 6.39\,cm$

• Tabular Presentation of Data:

  • Complete descriptive reporting requires structuring data tables with clear titles, labelled columns, raw values, calculated sample means ($\bar{x}$), and sample standard deviations ($s$).

Principles of Scientific Hypotheses

• Role of Hypotheses in the Scientific Method:

  • Following data collection, experiments evaluate formal statements regarding observed patterns.

  • Scientific testing requires establishing two distinct, opposing hypotheses.

• Alternative Hypothesis ($H_a$):

  • Definition: Proposes a specific relationship, effect, or difference between factors being investigated.

  • Criteria: Must be a clear declarative statement, testable, and falsifiable.

  • Notation: Symbolized as $H_a$.

  • Examples:

    • Walleye Study: Walleye length is larger in Oneida Lake than in Delta Lake ($H_a: \mu_{\text{Oneida}} \neq \mu_{\text{Delta}}$).

    • Bacterial Inhibition Study: Garlic extract creates a zone of bacterial inhibition that differs from that of water.

• Null Hypothesis ($H_0$):

  • Definition: Proposes that there is no relationship, no effect, and no difference between the experimental groups being compared.

  • Criteria: Must be testable and falsifiable.

  • Notation: Symbolized as $H_0$ (subscript zero/null).

  • Examples:

    • Walleye Study: There is no difference in the size of walleye between Oneida Lake and Delta Lake ($H_0: \mu_{\text{Oneida}} = \mu_{\text{Delta}}$).

    • Bacterial Inhibition Study: There is no difference in bacterial growth inhibition between garlic extract and water ($H_0: \mu_{\text{garlic}} = \mu_{\text{water}}$).

Statistical Hypothesis Testing and Decisions

• Foundational Testing Logic:

  • The null hypothesis ($H_0$) serves as the default statistical baseline.

  • $H_0$ is presumed true unless empirical evidence and statistical tests compel its rejection.

  • The primary objective of inferential statistical analysis is to determine whether to reject $H_0$.

• Mandatory Terminology for Conclusions:

  • "Reject the null hypothesis": Concluded when data demonstrates a statistically significant difference between groups.

  • "Fail to reject the null hypothesis": Concluded when empirical evidence is insufficient to demonstrate a significant difference.

  • Critical Distinction: "Failing to reject $H_0$" is not equivalent to proving $H_0$ true or correct; it merely indicates a lack of sufficient statistical evidence to disprove it.

• Overview of Common Inferential Statistical Tests:

  • Regression and Correlation: Evaluates relationships and predictive trends between continuous variables.

  • Two-Sample $t$-test: Compares means between two discrete sample groups (primary focus of biological sample analyses).

  • ANOVA (Analysis of Variance): Compares means across three or more sample groups.

  • Chi-Square ($\chi^2$) Test: Evaluates categorical data frequency distributions (frequently applied in genetics).

  • Logistic Regression: Models categorical or binary outcomes.

• General Experimental Workflow:

  1. Collect observational or experimental data (e.g., bacterial inhibition zones from garlic vs. water treatments).

  2. Calculate descriptive statistics (mean $\bar{x}$, variance $s^2$, standard deviation $s$).

  3. Compute test statistics (e.g., $t$-value) using computational software (Excel) to prevent manual calculation error.

  4. Extract probability values ($p$-value) to formulate formal conclusions.

Probability Values and Interpretation of Statistical Significance

• Probability Value ($p$-value):

  • Definition: The probability of obtaining an observed difference between groups purely due to random sampling chance, assuming the null hypothesis is true.

  • Range: $p$-values range continuously between $0$ and $1$.

• Standard Significance Threshold (Alpha Level, $\alpha$):

  • By scientific convention, the significance cutoff threshold is set at $\alpha = 0.05$ ($5\%$ probability of random chance).

• Decision Rules for Statistical Hypothesis Testing:

  • Condition 1: $p < 0.05$

    • Decision: Reject the null hypothesis ($H_0$).

    • Conclusion: There is a statistically significant difference between the group means being compared.

    • Interpretation: The likelihood that the observed difference occurred purely by random chance is less than $5\%$.

  • Condition 2: $p \ge 0.05$

    • Decision: Fail to reject the null hypothesis ($H_0$).

    • Conclusion: There is no statistically significant difference between the group means being compared.

    • Interpretation: Any observed numerical difference is attributable to random sampling variability.

• Hypothesis Formulation for $t$-Tests:

  • General Two-Sample Null Hypothesis: $H_0: \bar{x}_1 = \bar{x}_2$

  • General Two-Sample Alternative Hypothesis: $H_a: \bar{x}_1 \neq \bar{x}_2$

  • Walleye Specific Application:

    • $H_0$: The mean walleye length in Oneida Lake is equal to the mean walleye length in Delta Lake ($\bar{x}_{\text{Oneida}} = \bar{x}_{\text{Delta}}$).

    • $H_a$: The mean walleye length in Oneida Lake is not equal to the mean walleye length in Delta Lake ($\bar{x}_{\text{Oneida}} \neq \bar{x}_{\text{Delta}}$).

  • Bacterial Inhibition Application:

    • $H_0$: The mean bacterial inhibition zone around garlic is equal to the mean inhibition zone around water ($\bar{x}_{\text{garlic}} = \bar{x}_{\text{water}}$).

    • $H_a$: The mean bacterial inhibition zone around garlic is not equal to the mean inhibition zone around water ($\bar{x}_{\text{garlic}} \neq \bar{x}_{\text{water}}$).

• Sample Output Metrics:

  • Degrees of Freedom ($df$): Calculated based on total sample sizes ($n_1 + n_2 - 2$).

  • Calculated test statistic example: $t = 2.4$.

  • Final conclusion is driven by whether the calculated $t$-statistic yields a $p$-value falling below the $\alpha = 0.05$ critical threshold.