This preview shows pages 1–7. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: WHAT EFFECTS ARE SIGNIFICANT in 2 k Factorial Experiments? Model Y b b x b x b x N o Y = + + + + + + 1 1 2 2 12 12 2 ... ... ( , ) Significance of bs rests on estimate of Y 2 Which bs are significantly different than 0? Two Cases one replicate two or more replicates Determining Which Effects are Significant when there are two or more replicates Model & experiment Y b b x b x b x N o Y = + + + + + + 1 1 2 2 12 12 2 ... ... ( , ) m= 2 k treatments; n replicates; N = m n runs Estimate Y 2 with S p 2 1. Compute estimate of variance for treatment i = 1m ( 29 s i i i i n i i i n Y Y n where Y Y n 2 2 1 1 1 = = = = 2. Pool the m estimates S = S m p 2 i 2 i=1 m 2 Assumption Assume factors only affect the mean of the outcome not the variance In notation: for treatment i = 1m Y Y i 2 2 = 3 Example: a pooled estimated of y 2 2 2 Factorial Experiment with 2 replicates trt # I x 1 x 2 x 1 x 2 Y i1 Y i2 Y i S i 2 111 +1 .50 .46 .480 .0008 2 +111 .42 .37 .395 .00125 31 +11 .33 .30 .315 .00045 4 +1 +1 +1 .11 .16 .135 .00125 Compute S i 2 , a mini estimate of Y 2 ex i=1 [(.50  .48) 2 + (.46  .48) 2 ] / (21) = .0008 S 1 = .028 Compute S p 2 , an estimate for Y 2 S p 2 = [ .0008 + .00125 + .00045 + .00125]/4 4 = .0009375 and S p = .031 5 Use S p 2 to Compute Confidence Intervals for b = Y m= 2 k treatments; n replicates; N = m n runs estimate b Y = V Y N S N Y p ( ) = 2 2 estimate by 100(1 )% Confidence Interval for b N S t Y Y SD t Y p ) ( , 2 / ( ) ( , 2 / ( ) ( 1)m n = d.f....
View
Full
Document
 Fall '11
 Albin
 Ode

Click to edit the document details