1 Introduksi
1.1 Latar Belakang
Full factorial design $2^k$ adalah skema DOE yang mengeksekusi semua kombinasi level faktor ($2^k$ run untuk $k$ faktor dengan 2 level masing-masing). Keunggulan: estimasi main effects + semua interaction tanpa confounding (Fisher, 1935; Montgomery, 2017). Untuk $k = 3$ ($2^3 = 8$ run), full factorial bisa estimasi 1 grand mean + 3 main effects + 3 two-factor + 1 three-factor interaction = 8 parameter (saturated design).
Trade-off: jumlah run eksponensial dengan $k$. Untuk $k = 7$: $2^7 = 128$ run. Karena itu fractional factorial (di Screening Design Builder) sering dipilih untuk $k \geq 5$. Pada tahap confirmatory atau dalam screening ($k \leq 5$), full factorial paling informatif.
1.2 Tujuan Modul
Modul Factorial Design Builder di SQalytics ditujukan untuk:
- Membangun full factorial $2^k$ dengan $k = 2, 3, 4, 5$ faktor.
- Mendukung center points untuk lack-of-fit + curvature detection.
- Mendukung replicates untuk error estimation.
- Menampilkan alias-free estimation (vs fractional).
- Audiens: mahasiswa S2 industrial/food yang menjalankan DOE confirmatory, R&D dengan 3–5 faktor utama.
1.3 Posisi di Antara Alternatif
Pilih Factorial Design Builder untuk $k \leq 5$ faktor. Untuk $k \geq 6$ screening, pakai Screening Design Builder (fractional). Untuk multi-level optimasi, pakai RSM Studio (CCD, Box-Behnken). Untuk mixture proportion, pakai Mixture Design Explorer.
2 Metode
2.1 Dasar Teoretis
Coded levels: $-1$ (low) dan $+1$ (high). Untuk $k$ faktor, design matrix $N = 2^k$ baris × $k$ kolom.
Model effects untuk $2^3$:
$$ y = \beta_0 + \beta_1 x_1 + \beta_2 x_2 + \beta_3 x_3 + \beta_{12} x_1 x_2 + \beta_{13} x_1 x_3 + \beta_{23} x_2 x_3 + \beta_{123} x_1 x_2 x_3 + \varepsilon $$Main effect estimation (untuk orthogonal design):
Two-factor interaction effect:
$$ \text{Interaction}_{jk} = \frac{\sum_i y_i \cdot (x_{ij} \cdot x_{ik})}{N/2} $$Regression coefficient $\beta = \text{effect} / 2$.
Yates' algorithm (Yates, 1937) untuk efficient computation effects pada $2^k$ design (tabular hand calculation):
- Tabulasi response in standard order (Yates order).
- Pass-1: add pairs, subtract pairs.
- Pass-2: ulangi pada hasil pass-1.
- ... lanjut $k$ kali.
- Divide hasil akhir oleh $N/2$ untuk effects, $N$ untuk grand mean.
Sum of squares per effect:
$$ \text{SS}_{\text{effect}} = \frac{(\text{effect contrast})^2}{N \cdot r} $$dengan $r$ replicates.
Center points untuk curvature test:
$$ F_{\text{curvature}} = \frac{n_F n_C (\bar{y}_F - \bar{y}_C)^2}{(n_F + n_C) \text{MS}_E} $$dengan $\bar{y}_F$ mean factorial points, $\bar{y}_C$ mean center points, $n_F$ jumlah factorial run, $n_C$ jumlah center.
2.2 Persamaan Inti
Main effect: $\beta_j = \sum y_i x_{ij} / (N/2)$
Interaction: $\beta_{jk} = \sum y_i (x_{ij} x_{ik}) / (N/2)$
SS per effect: $\text{SS} = (\text{contrast})^2 / (N r)$
Curvature F-test: $F = n_F n_C (\bar{y}_F - \bar{y}_C)^2 / ((n_F + n_C) \text{MS}_E)$
2.3 Asumsi & Batas Validitas
| Asumsi | Konsekuensi jika dilanggar | Cara cek di SQalytics |
|---|---|---|
| Random sampling + randomization | Time bias | Auto-randomize |
| Orthogonality dipertahankan | Effect estimation bias | Check via coded matrix |
| Replicate cukup ($r \geq 2$) untuk error | $\sigma^2$ tidak teridentifikasi | Modul flag $r = 1$ |
| Center points $\geq 3$ untuk curvature | Curvature undetected | Default 4 CP |
| Linear assumption antara levels | Quadratic effect miss | Curvature test |
| Normal residuals untuk inferensi | t/F invalid | Pakai QQ-plot di ANOVA hasil |
3 Cara Kerja
3.1 Step-by-Step di SQalytics
- Buka
Factorial Design Builderdari domain Desain Eksperimen. - Daftarkan faktor ($k = 2, 3, 4, 5$): nama, satuan, low/high.
- Atur replicates ($r = 2$ default) dan center points (default 4).
- Aktifkan randomization.
- Klik
Generate Design. - Tinjau hasil:
- TabDesign Matrix— full $2^k \cdot r$ run + center points.
- TabRun Sheet— randomized order.
- TabPower Analysis— minimum detectable effect dengan $r$ replicates.
3.2 Template Tabel Input + Contoh Data Sintetis
Setup $2^3$ DOE ekstraksi (k = 3 factors, r = 2 replicates, 4 center points):
| Factor | Nama | Low (−) | High (+) | Center (0) |
|---|---|---|---|---|
| A | Temperature | 50 °C | 80 °C | 65 °C |
| B | Time | 15 min | 45 min | 30 min |
| C | Solvent ratio | 5 mL/g | 20 mL/g | 12.5 mL/g |
3.3 Contoh Luaran
Generated design (8 factorial + 8 replicate + 4 CP = 20 total runs):
| Run | A | B | C | Type |
|---|---|---|---|---|
| 1 | − | − | − | factorial |
| 2 | + | − | − | factorial |
| 3 | − | + | − | factorial |
| 4 | + | + | − | factorial |
| 5 | − | − | + | factorial |
| 6 | + | − | + | factorial |
| 7 | − | + | + | factorial |
| 8 | + | + | + | factorial |
| 9–16 | (replicates of 1–8) | replicate | ||
| 17–20 | 0 | 0 | 0 | center |
Power analysis:
| Metric | Value |
|---|---|
| $N$ total | 20 |
| Main effects | 3 (A, B, C) |
| Two-factor | 3 (AB, AC, BC) |
| Three-factor | 1 (ABC) |
| Center points | 4 |
| Degrees of freedom error | $20 - 8 - 4 = 8$ (with replicates) |
| Min detectable effect ($\alpha=0.05$, power=0.8) | $\sim 1.4\sigma$ |
y ~ A * B * C. Bila curvature signifikan, augment design ke Central Composite Design (CCD) di RSM Studio."RSM Studio (CCD atau Box-Behnken) untuk fit second-order polynomial dan optimasi presisi titik optimum.4 Kesimpulan
4.1 Relevansi Real-World
- Optimasi proses ekstraksi — Temperature × Time × Solvent untuk yield.
- Spray drying optimization — Inlet T × Outlet T × Feed rate untuk powder properties.
- Bread baking — Time × T × Hydration × Yeast untuk loaf volume.
- Yogurt fermentasi — Inoculum × T × Sugar × Time.
- Catalyst evaluation — 4 catalysts × 2 levels.
4.2 Where to Go from Here
⚙ Troubleshooting Cepat
i Riwayat Revisi
| Tanggal | Revisi | Penulis |
|---|---|---|
| 2026-05-12 | Migrasi MD v2 → HTML final dengan figure publikasi + caption Elsevier-style (W1 batch malam 12 Mei) | Claude |
| 2026-05-12 | Draft v2 publikasi (KaTeX $2^k$ + Yates + curvature F-test + APA Fisher/Montgomery/Yates) | Claude |
4 Referensi
- Fisher, R. A. (1935). The design of experiments. Oliver and Boyd.
- Montgomery, D. C. (2017). Design and analysis of experiments (9th ed.). John Wiley & Sons.
- Yates, F. (1937). The design and analysis of factorial experiments. Imperial Bureau of Soil Science.
- Box, G. E. P., Hunter, J. S., & Hunter, W. G. (2005). Statistics for experimenters: Design, innovation, and discovery (2nd ed.). John Wiley & Sons.
- Myers, R. H., Montgomery, D. C., & Anderson-Cook, C. M. (2016). Response surface methodology: Process and product optimization using designed experiments (4th ed.). John Wiley & Sons.