DoE CCD

[#1]

Hallo,

Ich habe eine Reihe von Experimenten mit DoE (Wirkungsflächendesign) erstellt. Nachdem ich die Experimente durchgeführt hatte, erhielt ich ein Pareto-Diagramm bei einem Signifikanzniveau von 0,02, in dem eine Referenzlinie (blaue Linie) eingezeichnet ist. Ich würde gerne wissen, bei welchem Wert meine Referenzlinie liegt? Kann ich den Wert berechnen lassen?

Vielen Danke

Grüße

[#2]

Hi @ZE : Welcome to the forum. It is the critical value from the t distribution. It depends on alpha (0.05 by default) and the error degrees of freedom (from the ANOVA table).  In JMP they are calculated (+/-), respectively, as shown below (where df = Error DF from ANOVA table).

-t Quantile( 0.975, df ), 
t Quantile( 0.975, df )

 This is explained in the Sorted Estimates help:

Sorted Estimates (jmp.com)

[#3]

Hi MRB3855

Thanks for the quick reply. I had 15 runs and set the alpha value to 0.02. Is there no way in the software to give me the value for the blue line (reference line)? I would like to create a diagram for my report and need the value for this line. In Minitab, I was given a value of 3.36 for the blue reference line.

Many thanks in advance.

[#4]

Hi @ZE Here ya go.

t Quantile( 0.01, 5 ) and  t Quantile( 0.99, 5 )

In general, t Quantile (alpha/2, Error df)  and t Quantile ( 1-alpha/2, Error df)

[#5]

Hi @MRB3855

I appreciate that. The main problem is that I don't get the same value as with Minitab (critical t-value = 3.36) and that makes me a bit unsure. I have 3 factors that have changed in three levels and occur 5 times each in the trials (15). So I have 14 DF and when I read the t-value at the 0.02 and 14 DF significance levels, I get 2.62.

[#6]

Hi @ZE : You have 15 total runs?  If so, you have 9 terms (3 main effects, 3 quadratic effects, and 3 pairwise interactions) in your model and that leaves only 5 df for Error. So 3.36 is correct; you have 14 df Total, but it is the Error df that matters. 

Or am I misunderstanding?

[#7]

Hi @MRB3855 

no, you're right. i'm sorry, i made a small logical mistake and always calculated with the total degree of freedom.