+ Reply to Thread
Results 1 to 3 of 3

Thread: Is this analysis (ANOVA and pairwise, with a potential cov.) statistically flawed?

  1. #1
    Points: 79, Level: 1
    Level completed: 58%, Points required for next Level: 21

    Posts
    3
    Thanks
    1
    Thanked 0 Times in 0 Posts

    Is this analysis (ANOVA and pairwise, with a potential cov.) statistically flawed?




    Hi Everyone,
    Below is an abbreviated adaption of my study. Subjects were to find three given types of faces - namely the Good, the Bad and the Ugly. In each trial, a Good, Bad or Ugly face was presented among three neutral ones, and the subject should move the mouse to the direction wherer the target face was. The goal was to test the effect of face type on reaction time, particularly, to see if people responded to Bad faces fastest.

    As a precautious measure, before the experiment we assessed the emotion intensity (EI) of each type of face - perhaps people would respond faster to a face showing intenser emotion. It turned out that Bad faces (our primary interest) showed higher EI than Ugly ones, but were comparable to Good ones.

    A one-way 3-level ANOVA showed that face type did influence reaction time. Then, follow-up pairwise comparisons showed Bad faces elicited faster response than Good ones, and than Ugly ones.

    My questions are as follows:
    1. As Bad faces showed higher EI than Ugly ones, is there any problem with interpreting the ANOVA as an effect of face type? (I think so because the F just tells you that at least two groups differ in reaction time, and this difference might be alternatively attributed to the difference in EI.)
    2. Given the rationale is flawed as in Q1, are the pairwise comparisons meaningful? (I think it is safe to say people responded faster to Bad faces than to Good ones since they were equal in EI, but not so safe with Bad vs. Ugly ones.)
    3. To test the effect of face type, is an ANCOVA necessary, with EI as a covariate?

    Any advice is appreciated.

    Thanks
    Meng

  2. #2
    Points: 3,506, Level: 37
    Level completed: 4%, Points required for next Level: 144

    Posts
    129
    Thanks
    3
    Thanked 26 Times in 26 Posts

    Re: Is this analysis (ANOVA and pairwise, with a potential cov.) statistically flawed

    I think you have answered your own questions 1 and 2! I agree an ANCOVA might be helpful. Even better might have been a selection of faces of each type that were matched for EI in the first place, to avoid this possible confound.

    I'm curious how you organized the data. Did each subject have one trial each with a good face, a bad face, and an ugly face? Or did each subject have many trials, and you averaged his or her responses to each face type? // in either case, was the analysis by repeated measures ANOVA (within subjects)?

  3. #3
    Points: 79, Level: 1
    Level completed: 58%, Points required for next Level: 21

    Posts
    3
    Thanks
    1
    Thanked 0 Times in 0 Posts

    Re: Is this analysis (ANOVA and pairwise, with a potential cov.) statistically flawed


    Thank you. It was a within-Ss design, and each subject had a sequence of trials, each with one of three types of face.

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats