Psychology 613
Data Analysis III
Prof. Bertram Malle
Spring 2007

Assignment 6

General Manova

In this assignment, you will analyze a multivariate one-way between-subjects design. The study tried to determine whether talking or touching by humans has an impact on plant growth, coloring, and durability. Thirty plants were examined in three conditions (no-intervention control, talking, touching). On various rating scales, independent judges (blind to the intervention) then rated each plant for growth, color saturation, and pliancy (squeeze and bend test).

The file /home4/bfmalle/613/assign5.dprg contains the data and command skeleton for the mainframe. The file /home4/bfmalle/613/assign5.dat contains the data only (with variable names in the first line). And here is an attempt to make the SPSS data file available that can directly be analyzed in SPSS for Windows/Mac. Press Control and click the mouse here to see the menu options; Choose "Download link to disk" or "Save Link As..." or the like.

  1. Examine the distributional properties (e.g., with EXAMINE) of the DVs and their intercorrelations. Can we expect Manova to be appropriate?

  2. After consulting the lecture handout and the SPSS chapter on the MANOVA procedure (especially pp. 105-119), fill in commands requesting the following features: (a) an appropriate contrast for the condition factor (justify your choice)
    (b) the observed means
    (c) the H matrix
    (d) a display of separate tests for each contrast, with effect sizes
    (e) the standardized discriminant function coefficients and loadings.     In a couple of sentences, explain which command yields which feature.

  3. Run the Manova and examine the cell means to get a sense of the patterns of data and describe in a few sentences what you see in these patterns.

  4. Inspect the H matrix. What can you say now about whether Manova is appropriate for this data set?

  5. Examine the omnibus main effect for condition. Describe the elements that make up the table called "Multivariate Tests of Significance" and identify the ones that may be reported. What have learned from the omnibus test?

  6. Then turn to the two contrasts (1st parameter, 2nd parameter). (a) What are the effect sizes for the contrasts? (b) Which variables contribute most to the contrasts? And (c) what do the contrasts indicate about the impact of our intervention?

  7. For clearer interpretation, run the same data through the DISCRIMINANT procedure. Examine the various plots the procedure offers (especially the all-groups scatterplot) to understand where the groups lie on the discriminant functions. Select the most useful plot, print it, and describe what it tells you.

  8. As usual, summarize your results in "journal format." Explain why MANOVA is more appropriate here than univariate ANOVA, explain your choice of contrasts, and describe the findings in your own words, reporting alongside the appropriate statistics (such as Pillai's V or Wilks' lambda, F-tests, and eta-squared).

  9. Page limit (including summary): 12