PSYC4005 Statistics Stream Assignment

A cognitive psychologist has been studying how short-term memory is influenced by three things: (1) Whether people are primed to use memory “intuitively” or “deliberately”, (2) their ability to deliberate about their own cognitive processes, and (3) cognitive style variables. She has set up an experiment where participants are randomly assigned to one of two conditions:

“prime”: -1 = intuition-primed 1 = deliberation-primed.

Prior to the experimental memorization task, they complete a test of deliberation ability (“ability”) and a battery of cognitive style measures (“v1-v17”).

Her dependent variables are how many nonsense words out of 30 the participant correctly recalls (“r”) and the amount of time taken in recalling them (RT).

  1. Conduct the appropriate data checking, cleaning and screening of the data for the RT variable, ability, prime and r variables. You need not do this for v1-v17.
  1. Variables v1-v17 are scales from three batteries of cognitive style measures. Conduct a principal components analysis (PCA) to assess the number of components that plausibly underlie all of these variables. Use a varimax rotation for the final solution. Save the 2component solution factor scores to use as DVs in the remainder of this assignment. In this part of your assignment, address the following questions:
    1. Would a one-component solution be an adequate summary of the entire collection of variables? Why or why not?
    2. Are there any items that could be dropped from the two component solution? Why or why not (and if so, which ones)?
  1. Two major hypotheses of this study are that deliberation ability will have a positive effect on response time, and intuition priming will result in shorter response times. Moreover, the researcher hypothesizes that deliberation ability will have a stronger effect in the deliberation-primed condition than in the intuition-primed condition. She also thinks that the cognitive style variables may affect response time, but they are not sure what kind of effect that will be. She would like you to conduct an ANCOVA on response time (or possibly a transformed version of it) that includes the priming condition, ability, and the two components from your PCA.
    1. Has homogeneity of regression been violated?
    2. What is your final model, and how did you arrive at it?
    3. Run the regression equivalent of your final model. Interpret the main and/or joint effects of the factor and the covariates. Evaluate the relative importance of these effects in terms of effect-sizes.
    4. Does ability have a significant effect on response time in both priming conditions?
    5. For your final model, compute 95% confidence intervals for the overall model etasquared and the partial eta-squared for the main effects.
  1. The researcher hypothesizes that the deliberation-priming condition and higher deliberation ability will result in a greater number of items correctly recalled (r) out of the 20 items to be memorized. Test her hypotheses with a logistic regression (testing for main effects only). Interpret any significant coefficients in odds-ratio terms.

Tip: In SPSS, use the Generalized Linear Models module for this, with the variables “r” and “n”.


  1. Address the questions in a results section of no more than 1000 words excluding title page and tables (a "word" is any nonblank character or string of characters separated from other characters or strings of characters by a space or punctuation mark — as measured by word count functions of word processors. A table or a graph counts as 1 word).
  2. The appendix must be no longer than 12 pages (and is worth 40% of the assignment marks). The appendix must be annotated. Write (legibly) or type comments on the page, explaining what conclusions you have drawn from the tests and plots. You will have to be selective about what output to include in the 12 page limit.

BACKGROUND. We are not going to take into account the psychological plausibility of models when marking your assignments.

You may use either SPSS or R (or both) for your analyses—Your mark does NOT depend on which package you use. You will need to provide the text of the actual commands (syntax) you ran in most analyses in the Appendix.