Stat 502 Homework 4 Assigned 10/25/07 Due 11/1/07 1. Clay surface roughness: A 3x3 factorial design was used to explore the effects of three sources of clay (I, II, III) and three ceramic molds (A, B, C) on surface roughness. Five replicates of each clay-mold combination were prepared and all 45 randomly placed within a furnace before firing. The response is a measure of surface roughness. The data are in the file claymold.rough.dat. (a) Examine these data graphically. What do you think about the data and effects of the two factors (clay source and ceramic molds) based just on this graphical examination of the data. Can you see any features of the data that you must pay attention to in formal statistical analysis? (b) Might a transformation be useful for these data? Why? (c) Write out the treatment effects model for the experiment and compute the least squares estimates of the parameters of this model. (This model may be on a transformed scale, depending on your answer to part (b).) (d) Compute the ANOVA table, including a column for the “Expected mean squares” (which you won’t get from R output). Is an additive model justified? Summarize your conclusions about the effects of clay source and ceramic molds. (e) Compute a 95% confidence interval for the contrast in cell means that you would use to contrast the difference between molds A and B using clay from source II with the difference between molds A and B using clay from source III. If you use the “Least Significant Difference” framework for hypothesis testing, do the results in part (d) authorize you to proceed with a test of the significance of this contrast? (f) Is there an optimal combination of clay source and mold if we are interested in minimizing the measure of surface roughness? That is, is there one combination of clay source and ceramic mold that has a mean score significantly less than all the other treatment combinations? Explain what statistical methodology you use to answer this question. 2. Plastic laminate cutting: The data below are production data for four different machines (1, 2, 3, 4) cutting four different types of laminated plastics (A, B, D, D). (a) Plot these data so that you can visualize the machine and plastic effects and comment. Do you observe any “bad values”? Explain. (b) There are no replications for the combinations of the “treatments” of Machine and Plastic, so consider an additive model. Then, for a range of values of the exponent λ (e.g., {-1, -.5, 0, 1/3, .5} evaluate the F statistic for the test of a machine effect on production using the power transformation Yλ where Y denotes production. Similarly, evaluate the F statistic for the effect of the type of plastic over these same values of λ. Does this help you determine an appropriate transformation? Choose your preferred transformation and plot the usual diagnostics using the residuals from the fit of the additive model on the transformed scale. What conclusions do you draw? Plastic 1 2 3 4 A 1271 1440 612 605 Machines B C 4003 1022 1651 372 1664 829 2001 258 D 2643 5108 1944 2607