Index: The Book of Statistical Proofs ▷ Statistical Models ▷ Univariate normal data ▷ Analysis of variance ▷ Sums of squares in two-way ANOVA

Theorem: Given two-way analysis of variance,

\[\label{eq:anova2} y_{ijk} = \mu + \alpha_i + \beta_j + \gamma_{ij} + \varepsilon_{ijk}, \; \varepsilon_{ijk} \overset{\mathrm{i.i.d.}}{\sim} \mathcal{N}(0, \sigma^2)\]

sums of squares can be partitioned as follows

\[\label{eq:anova2-pss} \mathrm{SS}_\mathrm{tot} = \mathrm{SS}_{A} + \mathrm{SS}_{B} + \mathrm{SS}_{A \times B} + \mathrm{SS}_\mathrm{res}\]

where $\mathrm{SS}_\mathrm{tot}$ is the total sum of squares, $\mathrm{SS}_{A}$, $\mathrm{SS}_{B}$ and $\mathrm{SS}_{A \times B}$ are treatment and interaction sum of squares (summing into the explained sum of squares) and $\mathrm{SS}_\mathrm{res}$ is the residual sum of squares.

Proof: The total sum of squares for two-way ANOVA is given by

\[\label{eq:anova2-tss} \mathrm{SS}_\mathrm{tot} = \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (y_{ijk} - \bar{y}_{\bullet \bullet \bullet})^2\]

where $\bar{y}_{\bullet \bullet \bullet}$ is the mean across all values $y_{ijk}$. This can be rewritten as

\[\label{eq:anova2-pss-s1} \begin{split} \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (y_{ijk} - \bar{y}_{\bullet \bullet \bullet})^2 = \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} & \left[ (y_{ijk} - \bar{y}_{i j \bullet}) + (\bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet \bullet \bullet}) + (\bar{y}_{\bullet j \bullet} - \bar{y}_{\bullet \bullet \bullet}) + \right. \\ \\ & \left. (\bar{y}_{i j \bullet} - \bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet j \bullet} + \bar{y}_{\bullet \bullet \bullet}) \right]^2 \\ \end{split}\]

It can be shown that the following sums are all zero:

\[\label{eq:anova2-pss-s2} \begin{split} \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (y_{ijk} - \bar{y}_{i j \bullet}) &= 0 \\ \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (\bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet \bullet \bullet}) &= 0 \\ \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (\bar{y}_{\bullet j \bullet} - \bar{y}_{\bullet \bullet \bullet}) &= 0 \\ \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (\bar{y}_{i j \bullet} - \bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet j \bullet} + \bar{y}_{\bullet \bullet \bullet}) &= 0 \; . \end{split}\]

This means that the sum in \eqref{eq:anova2-pss-s1} reduces to

\[\label{eq:anova2-pss-s3} \begin{split} \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (y_{ijk} - \bar{y}_{\bullet \bullet \bullet})^2 = \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} & \left[ (y_{ijk} - \bar{y}_{i j \bullet})^2 + (\bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet \bullet \bullet})^2 + (\bar{y}_{\bullet j \bullet} - \bar{y}_{\bullet \bullet \bullet})^2 + \right. \\ \\ & \left. (\bar{y}_{i j \bullet} - \bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet j \bullet} + \bar{y}_{\bullet \bullet \bullet})^2 \right] \\ = \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} & (y_{ijk} - \bar{y}_{i j \bullet})^2 + \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (\bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet \bullet \bullet})^2 + \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (\bar{y}_{\bullet j \bullet} - \bar{y}_{\bullet \bullet \bullet})^2 \\ \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} & (\bar{y}_{i j \bullet} - \bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet j \bullet} + \bar{y}_{\bullet \bullet \bullet})^2 \; . \end{split}\]

With the treatment sums of squares

\[\label{eq:anova2-trss} \begin{split} \mathrm{SS}_A &= \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (\bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet \bullet \bullet})^2 \\ \mathrm{SS}_B &= \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (\bar{y}_{\bullet j \bullet} - \bar{y}_{\bullet \bullet \bullet})^2 \; , \end{split}\]

the interaction sum of squares

\[\label{eq:anova2-iass} \mathrm{SS}_\mathrm{A \times B} = \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (\bar{y}_{i j \bullet} - \bar{y}_{i \bullet \bullet} - \bar{y}_{\bullet j \bullet} + \bar{y}_{\bullet \bullet \bullet})^2\]

and the residual sum of squares for two-way ANOVA

\[\label{eq:anova2-rss} \mathrm{SS}_\mathrm{res} = \sum_{i=1}^{a} \sum_{j=1}^{b} \sum_{k=1}^{n_{ij}} (y_{ijk} - \bar{y}_{i j \bullet})^2 \; ,\]

we finally have:

\[\label{eq:anova2-pss-qed} \mathrm{SS}_\mathrm{tot} = \mathrm{SS}_{A} + \mathrm{SS}_{B} + \mathrm{SS}_{A \times B} + \mathrm{SS}_\mathrm{res} \; .\]
∎
Sources:

Metadata: ID: P379 | shortcut: anova2-pss | author: JoramSoch | date: 2022-11-16, 16:01.