Dear sir,

I am using Design Expert 10 software for analysis. Can you please help me in identifying significant effects from half normal plot?

Thanks in advance

Prasenjit

yes, half normal plot can be used.

you may contact me for details about its analysis if required.

[email protected]

Dear madam,

Thank you very much for your reply.

I am attaching the excel file containing the output of the analyses, half normal plot and pareto chart. I have 16 factors to screen and factors R, S and T are dummy factors. Are you able to download the file? If you cannot, please let me know.

If I select the six effects (L, J, P, F, M, B, although J looks to be on the line), the p values of all the parameters are coming less than 0.05, i.e., all the selected parameters are significant at 95% confidence level. If I deselect any of them or add any extra parameter, then the p-value for some of the parameters are coming out to be more than 0.05.

The factor L, that is almost on the line is the most significant one. If I don't select that effect, then warning message pops up "Largest effect not selected". Generally the points lying off the fitted line are the most significant ones. In that case it is opposite.

I deselected the effect J falling very close to the fitted line. Then in the ANOVA, again the p values came to be more than 0.05 for some of the selected parameters, which means that if I  select 5 parameters and out of them 2 parameter's p values are coming to be more than 0.05, then I'll have to work with only 3 parameters. 

I tried to find the significant effects based on AICc, BIC, p-value and Adj. R-squared based algorithms inbuilt in the Design Expert Software (by selecting all the factors and then running the algorithm). Outcome was that the six factors that I was telling you (B, F, L, M, J and P) are to be included in the model.

I am looking forward for your kind reply.

Thanks in advance.

Prasenjit Ghosh

Dear madam,

Yes. Sure. You take your time. Thank you very much for your kind reply.

Thanks and regards

Prasenjit Ghosh

Dear Ghosh,

The following link will help the solution of your problems.

https://www.isixsigma.com/tools-templates/design-of-experiments-doe/when-and-how-to-use-plackett-burman-experimental-design/

Dear Muzaffer Balbaşı (Sir),

Thank you for your reply.

I have gone through the link. I am using Design Expert 10 software. Previously I had run my design data using Minitab 17, but the results did not match with that obtained using Design Expert Software. I will look into more details about the facts mentioned in the link.

Thanks and regards

Prasenjit

Dear madam,

Thank so much for your effort and reply to my answer. I have gone through your analyses files. I have few doubts:

1. On what basis/logic  did you select the factors A, B, F, G, J, L, M, P and Q in the half normal plot for performing ANOVA test? Is it based on their % contribution values as shown in the Effects list?

2. f I deselect the term Q from the half normal distribution, then L, J, P, F, M and B are the significant terms as per the Pareto chart.

3. In addition, if I deselect the terms Q, A and G, then also the significant terms come out to be L, J, P, F, M and B.

4. If I select all the parameters except the dummy variables, R, S and T, then J and L are coming out to be the only two significant effects as per the ANOVA at 95% confidence level.

I am not understanding the selection criteria of significant effects.

Please explain.

Thanks in advance

Prasenjit Ghosh

Dear madam,

In addition to my previous post, I want to mention that my next experimental design is factorial design. In that case, I have restrict my design variables up to maximum of 4 (so that I have 2^4 = 16 runs), due to lack of resources and time. In that case, will it be logical to select less numbers of factors? I am planning on selecting the factors L and J (since their t-values are above Bonferroni limit. From the group of probably significant effects, I have to select 1 or max. 2. If I select the effect F from the group (since its % contribution is 11.4 (higher than M and B, 9.77 each) will be justifiable? If I select one more variable, then I can select M and then automatically I have also to choose B since their % contribution values are equal. Is it ok to choose L, J and F? Please explain.

Thanks in advance

Prasenjit Ghosh

Dear madam,

Thank you for your kind reply.

I want to mention that my next experimental design plan is factorial design not fractional factorial design. 

For factorial design, I have restrict my design variables up to maximum numbers of 4 (so that I have 2^4 = 16 runs), due to lack of resources and time. In that case, will it be logical to select less numbers of factors? Which 3 or 4 factors should I choose for the purpose and how can I justify their selection? Another question....what is the role of the term % contribution in Analysis--> R1 % degradation--> Effects--> Numeric--> % contribution (is mentioned in the right most column)? Is it also considered for deciding the significant effects in PB design?

The parameters of the Plackett Burman Design are:

A: KH2PO4; B: K2HPO4; C: Na2HPO4.2H2O; D: CoCl2.2H2O; E: H3BO3; F: Na2MoO4.2H2O; G: CuCl2.2H2O; H: ZnCl2; J: NiCl2.6H2O; K: FeCl2.4H2O; L: MnCl2.4H2O; M: Na2SO4; N: NH4Cl; O: MgSO4.7H2O; P: CaCl2.2H2O; Q: FeCl3; R: Dummy 1; S: Dummy 2 and T: Dummy 3.

A, B and C are the major nutrients, combinedly added to prepare phosphate buffer; D to M are trace elements and N to Q are other elements added in the mineral media.

Thank you in advance

Prasenjit Ghosh

Dear madam,

Thank you very much for your reply.

I'll be doing central composite design using RSM which comes under factorial design. Actually we don't have dedicated shaker for doing only my experiments, they are in common, so I have to maintain 37 0C and 120 rpm. I can't afford to optimize those parameters. I am attaching here the concentration ranges that I had chosen for performing the studies. My experimental response is the % degradation of a PAH compound.

Thank you in advance

Prasenjit