Null Hypothesis (H₀): States there is no real difference or association. (e.g., "The new drug has no effect on parasite load compared to a placebo"). Any difference is due to chance.
Alternative Hypothesis (H₁): States there is a real difference. (e.g., "The drug reduces parasite load").
Decision Errors
- Type I Error (α): Rejecting H₀ when it is actually true. (False Positive). p < 0.05 is the standard probability threshold.
- Type II Error (β): Failing to reject H₀ when it is actually false. (False Negative).
- Power of the Test (1 - β): Probability of correctly detecting an effect that actually exists.
Parametric Tests
Assume data follows a normal distribution (Gaussian), variances are equal (homoscedasticity), and data is continuous.
e.g., t-test, F-test, ANOVA
Non-Parametric Tests
"Distribution-free" tests used when data is skewed, sample sizes are tiny, or data is nominal/ordinal.
e.g., Chi-square, Mann-Whitney U, Kruskal-Wallis
C. The Four Heavyweight Tests
Used when population standard deviation is unknown.
One-Sample t-test:
Compares a sample mean to a known population mean.
Independent/Two-Sample t-test:
Compares means of two distinct groups (e.g., wild vs captive frogs).
Paired t-test:
Compares two means from the same individuals measured twice (e.g., weight before/after diet).
F = s₁² / s₂²
Note: s₁² > s₂², meaning the larger variance is always the numerator (F ≥ 1).
Crucial Application:
Checks the assumption of equal variances before a t-test and forms the backbone of ANOVA.
Why not multiple t-tests?
Running 6 t-tests for 4 groups makes the cumulative Type I error skyrocket to ~26.5%. ANOVA controls this by testing all means at once.
Mechanism: ANOVA splits variation into:
- Between Groups: Measures the effect of treatments.
- Within Groups (Error): Measures random individual differences.
1. Goodness of Fit: Tests if observed ratios match theoretical Mendelian ratios (e.g., 3:1).
2. Test of Independence: Checks if two categorical variables are linked (e.g., disease vs sex).
Degrees of Freedom (df):
• Goodness of Fit: df = k - 1
• Contingency Table: df = (r - 1)(c - 1)
Scenario:
A genetics researcher crosses two heterozygous fruit flies (AaBb × AaBb) and counts the phenotypes of 400 offspring to determine whether the two genes are assorting independently according to standard Mendelian laws (9:3:3:1).
Question 1:
Which specific test of significance should the researcher use to analyze this data?
Question 2:
How many degrees of freedom (df) will be used when looking up the critical value in the statistical tables?
Hint: Categorical count data vs. Ratio.