Discovering Statistics

Adjusted error bars for repeated measures designs

Viewing 1 post (of 1 total)
  • Author
    Posts
  • #554
    Gavin Revie
    Member

    I’ve been slowly working my way through Discovering Statistics Using R.  I just got to chapter 9 – comparing means.

    I’ve been reading the section on creating adjusted error bars for repeated measures designs.  I was completely ignorant of this (as are a lot of people as far as I can see) and I’ve been making my error bars look worse than they had to for years.  So first of all I’d like to say thank you to Andy Field for drawing this to my attention.

    Now, in the book he cites Loftus and Masson 1994 before launching into an explanation of the adjustment process.  However the method he describes appears to be that of Cousineau 2005.  The Loftus method involves running an ANOVA first.  After a little reading I found that Cousineau 2005 has been criticised for making the error bars slightly too small, and a correction was proposed by Morey 2008.  This in turn has been criticized by Franz and Loftus 2012.

    Given that I was new to all of this all these competing methods are rather making my head spin.  Can anyone help me unpick them?

    Bear in mind that this is exclusively for the purposes of creating error bars on graphs.  Obviously the actual analysis, be it t-test or ANOVA or whatever, will be done in the normal way that corrects for a repeated measures design.

    Part of me wonders if all of this angst about error bars isn’t slightly OTT and whether I should just use the method described in DSUR.  But then, if there are reasonably easy R implementations of the better methods, obviously it would be good to use them.  Can anyone point me to such resources, or offer any input on this whole conundrum?

    Thank you for any help you can provide.

Viewing 1 post (of 1 total)
  • You must be logged in to reply to this topic.