Evaluasi Perbandingan Efisiensi Rancangan Percobaan Klasik dan Rancangan Optimal dalam Mengestimasi Efek Perlakuan

Abstract

This study evaluates the comparison between classical experimental designs and optimal designs in efficiently estimating treatment effects, with a focus on their ability to produce accurate parameter estimates and minimize variance. Classical experimental designs, such as the factorial Completely Randomized Design (CRD), are widely applied due to their simplicity and ease of implementation. However, their effectiveness may decline in situations involving multiple factors or limited resources. In contrast, optimal designs, such as the A-optimal design, are specifically constructed to minimize the variance of parameter estimates, thereby providing more precise information using fewer experimental units. The comparison is conducted in the context of evaluating the interaction effect between fertilizer dosage and plant spacing on rice productivity. The assessment is carried out through simulation and analysis of the Fisher Information Matrix to measure the design's statistical efficiency. The results show that the A-optimal design provides more accurate and efficient estimates, especially in accommodating complex factor interactions. Therefore, the A-optimal approach can be employed to improve suboptimal experimental designs. Furthermore, the choice of an experimental design should be guided by the study objectives, model complexity, and available resources to ensure optimal experimental outcomes.