DOE

 Week 12 Blog - Design of Experiment (Case 1)

For case 1, I am using EXCEL to determine the amount of bullets (unpopped kernels) in grams remain at the bottom of the bag. These 3 factors [diameter (A), microwaving time (B) and power (C)] will be tested with 100 grams of corn. Here is the data provided: 

First I find the average when A+, A-, B+,  B-, C+ and C- and compare them in a line graph. This is the FULL Factorial Analysis.

Gradient of each line will represent our significant the change is. Therefore based on the graph above, the sequence will be C > B >A in terms of its effect to decrease the number of unpopped kernels. Increasing the power from 75% to 100% decreases the number of bullets the most followed by increasing microwaving time from 4mins to 6 mins and then increasing diameter of bowl from 10cm to 15cm.

Interaction effect is when two factors interact with each other, the effect of one factor on the response variable is different at different levels of the other factor. Now this is the 3 graph for Interaction Effect for FULL Factorial Design (AxB, AxC, BxC):

For A x B, one line has a positive gradient and the other has a negative gradient but both line is similar in terms of its magnitude so there is a small interaction between A and B.

For A x C, both lines are almost parrallel with each other so there is a smaller interaction between A and C compared to A and B.

For B x C, both lines have a negative gradient and of very different values so there is a significant interaction between B and C.

 For my FRACTIONAL factorial analysis, I chose runs 2, 3, 4, 5 because all of the factors occur both low and high the same number of times (orthogonal). Likewise with full factorial analysis, I compare the factors using the 4 runs.

However, the ranking is different. For my FRACTIONAL factorial ranking is B=C, A from most significant to least significant. Based on the graph, the Power from 75% to 100% decreases the mass of bullets the most followed by change in diameter of bowl from 10cm to 15cm. There is an anomaly where increasing microwaving time from 4 minutes to 6 minutes increases the mass of bullets, therefore more replications should be run so that all the data will be more reliable.

File to Data Analysis (Graphs bugged when transferring from Excel to Google Sheets)