Exercises 411whether different doses would have different effects. Data that have been generated to
loosely mimic the results of Stone et al. are given below, where the dependent variable is
the latency to enter the dark chamber.
Glucose Level in mg/kg
0 1 10 100 250 500
295 129 393 653 379 521
287 248 484 732 530 241
91 350 308 570 364 162
260 278 112 434 385 197
193 150 132 690 355 156
52 195 414 679 558 384a. Plot these data using both the actual dosage, and the values 1, 2, 3, 4, 5, 6 as the values
of X.
b. Run a trend analysis using SPSS Oneway, if available, with the actual dosage as the
independent variable.
c. Repeat part b) using the 1, 2, 3, 4, 5, 6 coding as the independent variable.
d. Interpret your results. How might these results have something to say to students who
stay up all night studying for an exam?
e. Why might you, or Stone et al., prefer one coding system over another?
12.26 Using the data from Exercise 12.1, compute confidence interval for the first comparison (con-
trast) described in that question. Interpret your answer. (If you use SPSS, use the Compare
Means/One-Way ANOVAprocedure, which allows you to specify coefficients.)
12.27 Using the data from Exercise 12.1, compute effect sizes on all of the contrasts that you ran
with that question. How would you interpret these effect sizes? Why are these called stan-
dardized effect sizes, and what would an unstandardized effect size be?
12.28 Write up a brief report of the results computed for Exercises 12.1, 12.26, and 12.27.
12.29 Using the data from Exercise 11.27, perform the appropriate test(s) to draw meaningful con-
clusions from the study by Davey et al. (2003).
12.30 In Exercise 11.8 we considered a study by Foa et al. concerning therapy for victims of rape.
The raw data can be found on the Web site at Ex12.30.dat. Apply the Benjamini and
Hochberg LSU procedure to these data.Discussion Questions
12.31Students often have difficulty seeing why a priori and post hoc tests have different family-
wise error rates. Make up an example (not necessarily from statistics) that would help to ex-
plain the difference to others.
12.32 Find an example in the research literature of a study that used at least five different condi-
tions, and create a data set that might have come from this experiment. Apply several of the
techniques we have discussed, justifying their use, and interpret the results. (You would
never apply several different techniques to a set of data except for an example such as this.
Hint: You can generate data with a given mean and variance by taking any set of numbers
[make them at least unimodal and symmetrical], standardizing them, multiplying the stan-
dard scores by the desired standard deviation, and then adding the desired mean to the
result. Do this for each group separatelyand you will have your data.)