Estimating Missing Data in Unreplicated 2k Factorial Experiments

Main Article Content

Kanidta Laphaphong
Vichit Lorchirachoonkul
Jirawan Jitthavech


In this study, two estimation methods for one missing data in unreplicated 2k factorial experiments are proposed. The first method is minimizing sum squared the highest order interaction and the second one is based on the change proportion. From the comparison of 3 measurement criteria: the absolute percentage relative error of the estimate, sum squared and estimates of factorial effects, by using 22, 23 and 24 factorial experiments, it can be concluded that the change proportion method gives a better estimate than the method of minimizing sum squared the highest order interaction.


Download data is not yet available.

Article Details

Applied Science Research Articles


[1] J. Jitthavech, Design and Analysis of Experiment, Bangkok: Thaipattana Raiwan Press Company Limited, 2009 (in Thai).

[2] P. Suwattatthee, Sample Surveys: Sampling Design and Analysis, Bangkok: WVO Office of Printing Mill The War Veterans Organization of Thailand Under Royal Patronage of His Majesty the King, 2009 (in Thai).

[3] D. Montgomery, Design and Analysis of Experiments,5th ed., John Wiley and Sons, 2001.

[4] C. F. Jeff Wu and M. Hamada, Experiments: Planning Analysis and Parameter Design Optimization, John Wiley and Sons, 2000.

[5] N. Draper and D. Stoneman, “Estimating missing values in un-replicated two-leval factorial and fractional designs,” Journal of Biometrics, vol. 20, no. 3, pp. 443–458, 1964.

[6] M. Qumsiyeh and K. Kiechner, “Estimation methods for missing data in un-replicated 2k factorial and 2k-p fractional factorial designs,” Journal of Statistics: Advances in Theory and Applications, vol. 5, no. 2, pp. 131–147, 2011.

[7] Science Division, Dictionary of Statistics, Bangkok: Office of the Royal Society, 2015 (in Thai).

[8] G. Box, W. Hunter and J. Hunter, Statistics for Experiments, John Wiley and Sons, 1978.