• Null Hypothesis ($H_0$): All population means are exactly equal ($\mu_1 = \mu_2 = \dots = \mu_k$). The factor has no effect.
• Alternative Hypothesis ($H_1$): At least one population mean is different from the others. (Note: It does not mean they are all different from each other).

The results are summarized in an ANOVA table. The test statistic is the $F$-ratio:

If $H_0$ is true, the between-group and within-group variances should be roughly equal, yielding an $F$-ratio near 1. A large $F$-ratio (and a correspondingly small P-value) provides evidence to reject $H_0$.

The core of ANOVA is dividing the total variability of the data (Total Sum of Squares, $SST$) into two parts: the variability due to differences between the group means (Treatment Sum of Squares, $SSTr$) and the random variability within the groups (Error Sum of Squares, $SSE$). Thus, $SST = SSTr + SSE$.

• Independence: The observations within each group, and between groups, are independent. Random sampling/assignment is crucial.
• Normality: The populations from which the samples are drawn are normally distributed. ANOVA is generally robust to mild violations if sample sizes are equal.
• Homogeneity of Variances (Homoscedasticity): The variances of the different populations must be equal ($\sigma_1^2 = \sigma_2^2 = \dots = \sigma_k^2$). Tested using Levene's Test or Bartlett's Test. If violated, a transformation (like taking the log of the data) or a non-parametric alternative (Kruskal-Wallis) is required.

• Fisher's Least Significant Difference (LSD): The most liberal test (highest power to detect differences, but highest risk of Type I error). It is essentially a series of t-tests performed only after a significant ANOVA.
• Tukey's Honestly Significant Difference (HSD): The gold standard for comparing all possible pairs of means. It rigorously controls the family-wise error rate (the probability of making any Type I error across all comparisons).
• Dunnett's Test: Used specifically when comparing several treatments against a single "control" group, rather than comparing every group against every other group.

In RCBD, experimental units are grouped into homogeneous "blocks" (e.g., testing each cement type on Machine 1, then Machine 2, etc.).

• The total variability is partitioned into three sources: Treatments, Blocks, and Error.
• By isolating the variability caused by the blocks, the remaining random error (MSE) is significantly reduced.
• A smaller MSE denominator increases the $F$-ratio for the main treatment, making the test much more powerful (sensitive) at detecting true differences between the cement types.