##
Perancangan Eksperimen [8]
*March 27, 2009*

*Posted by desrinda in Perancangan Eksperimen.*

add a comment

add a comment

Starting this 8th week, Bu Arian (our professor in Statictics Class, first term) has started teaching us again.

Taken from Chapter 5-4 of our textbook (*Design and Analysis of Experiments*, Douglas C. Montgomery, John Wiley & Sons, 2005), we were learning about a sampel of experiment in three two-way tables to compute ANOVA.

##
Perancangan Eksperimen [7]
*March 20, 2009*

*Posted by desrinda in Perancangan Eksperimen.*

add a comment

add a comment

Last class by Bu Isti: Latin Square Design.

##
Perancangan Eksperimen [6]
*March 14, 2009*

*Posted by desrinda in Perancangan Eksperimen.*

add a comment

add a comment

**FACTORIAL DESIGN**

– Mean Square Calculations

– The F Test Statistics

Two-factor model can be analysis by:

– main effect plots

– interaction plots

– normal probability plots

– other residual analysis

##
Perancangan Eksperimen [5]
*March 7, 2009*

*Posted by desrinda in Perancangan Eksperimen.*

add a comment

add a comment

**BLOCKING IN ONE-FACTOR ANOVA**

A nuisance factor is a factor that probably has some effect on the response but not considered by the experimenter. However, the variability it transmits to the response needs to be minimized.

Types of nuisance factor:

1. Unknown and uncontrollable => RANDOMIZATION

2. Known but uncontrollable => ANALYSIS OF COVARIANCE (ANCOVA)

3. Known and controllable => BLOCKING TECHNIQUE

Variability arising from nuisance factor can impact the results of an experiment. Therefore, experimenter should apply randomized complete block design (RCBD, Chapter 4 of *Design and Analysis of Experiments*, Douglas C. Montgomery, John Wiley & Sons, 2005).

##
Perancangan Eksperimen [4]
*February 27, 2009*

*Posted by desrinda in Perancangan Eksperimen.*

add a comment

add a comment

The **Analysis of Variance** (ANOVA) is the appropriate analysis “engine” for experiments with more than two factors to be taken into calculation.

**SST = SSA + SSW**

SST: Total sum of squares (total variations)

SSA: Among-group variations (variation due to factor)

SSW: Within-group variations (variation due to random sampling)

*Assigment*: Find out about corelation between **F-ratio** and **chi-square**.

##
Perancangan Eksperimen [3]
*February 22, 2009*

*Posted by desrinda in Perancangan Eksperimen.*

add a comment

add a comment

During the class, we were studying basic statistics terminologies. All of my classmates have enjoyed Bu Isti’s explanations, she’d great passion in teaching then we could get more understanding on terms previously learnt (Statistik Industri).

Data presentation:

– nominal (has no mean, median, etc)

– ordinal (numbers in order)

– interval

– ratio (can apply math operators)

**VARIATION**

= Data spread or diversion, consists of range + variance + standard deviation.

For continuous data, it is necessary to transform it to presented as standardized normal distribution.

##
Perancangan Eksperimen [2]
*February 15, 2009*

*Posted by desrinda in Perancangan Eksperimen.*

add a comment

add a comment

Review on basic statistical points to be considered while designing experiments: Case study at Telkomsel.

##
Perancangan Eksperimen [1]
*February 8, 2009*

*Posted by desrinda in Perancangan Eksperimen.*

add a comment

add a comment

Masa perkuliahan di semester kedua diawali dengan kelas Perancangan Eksperimen (DoE, Design of Experiment), diajar oleh Ibu Ir.Isti Surjandari, MT, MA, PhD. Menurut silabus, melalui kuliah ini pesertanya diharapkan akan mampu mengorganisir pengumpulan, pengolahan, dan penganalisaan data dengan baik & benar secara statistik maupun engineering (rekayasa) dalam melakukan rancangan percobaan yang bertujuan untuk pengambilan keputusan.

Textbook: *Design and Analysis of Experiments*, Douglas C. Montgomery, John Wiley & Sons, 2005.

**BASIC PRINCIPLES**

Principle of parsimony (statistics) = The scientific principle that things are usually connected in the simplest or most economical way.

KISS = Keep It Statistically Simple, or “Keep it simple, Stupid!”

Any process can be modelled into simple flow chart:

X = controllable factors, anything that affect output

Y = uncontrollable factors, should be blocked (can’t be eliminated)

Outputs should be measurable, i.e.

– Quantity of received customer’s complaints (service industry)

– Quantity of rejected item (manufacturing)

– Food taste (in scale from 1 to 5)

– etc.

Point of departure must be set by defining hypothesis to be tested.

Ho = No relationship between controllable factors and outputs

Ha = There is relationship between controllable factors and outputs

Purpose of analysis are vary: For understanding, explanation, prediction, or control. In general, to decide which factors can increase efficiency.

The strategy of experiments:

– Best guess

– One-Factor-at-A-Time (OFAT)

– Statistically designed experiments = DoE, taking summaries of result of interaction between factors

Dr. Leonard Lye taught DoE (STAT-EASE, Statistics Made Easy) using DoE golfer as sample. The software was Minitab.

Based on Fisher’s factorial design:

– 2 factors, 2 level each = 2 x 2 = 4 runs

– 3 factors, 2 level each = 2 x 2 x 2 = 8 runs

– and so on

If not all of run is taken into experiment, then it’s called fractional factorial design.

Steps:

1. Statement of problem (e.g. using fishbone diagram)

2. Design of experiment (method, hypothesis, etc.)

3. Analysis

Six Sigma: DMAIC = Define, Measure, Analysis, Improve, Control.

Statistical DoE Principles: