{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

# power - to loop on s02 since L is ﬁxed at 1 eliminate...

This preview shows page 1. Sign up to view the full content.

Power Analysis Exercise I would like you to work in large groups in discussing coding changes; you will want to separate into groups of 2 or 3 when running the code. Hand in informal responses to questions. Suppose you have a Completely Randomized Design with 4 levels: 3 treatments and 1 control. In the analysis, you are particulary interested in comparing the control mean against the average of the treatment means: H o : μ 1 + μ 2 + μ 3 3 - μ 4 = 0 1. What are the contrast coefficients? Substitute your answer into the formula for the noncentrality parameter for a contrast and simplify. 2. Suppose σ 2 = 4 and you would like to detect an alternative contrast of 1; use this information to rewrite the noncentrality parameter as a simple function of n . Now modify power.sas in order to conduct a power analysis. Note that you do not need
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: to loop on s02 since L is ﬁxed at 1; eliminate this loop. Based on your modiﬁed code, construct a scatterplot of the power as a function of n; what range of n values gives you good power to detect L = 1? 3. With σ 2 = 4, modify power.sas to compute power for a range of choices of L and n ; it helps if you rewrite the noncentrality parameter nc as a function of L . The output ﬁle should contain the power for a grid of values of n and L . 4. Produce a power contour plot in Minitab (see homepage for using Minitab’s contour plot). You may have to manipulate your choice of n in order to obtain a satisfactory plot. What are some choices of (L,n) for the .80 power contour? 1...
View Full Document

{[ snackBarMessage ]}