Print

Print
Hi Kai,

With one less factor, it's simpler, and the design can proceed as a simple 2x2 repeated measures. Please, see here:
https://dl.dropboxusercontent.com/u/2785709/outbox/mailinglist/design_kai.ods

Even though there are twins, it seems you only have MZ and DZ pairs, and so, differently than the HCP, that has more complicated sibships. This means it isn't strictly necessary to define, by hand, the multi-level exchangeability blocks. A simple column, with whole and/or within-block permutation is sufficient (internally PALM will create the multi-level structure, but it doesn't have to be specified by hand).

In the updated design, design1 is for H vs. L, and for the interaction of performance with zygosity, whereas design2 is for differences associated with zigosity while taking differences in performance into account. The design1 uses within-block permutation, whereas design2 uses while-block.

Hope this helps.

All the best,

Anderson


On 10 December 2015 at 00:38, Kai Wang <[log in to unmask]> wrote:
Hi Anderson,

Thank you so much for your time!

However, maybe I didn’t state it clearly, but we are talking about a higher level model, whose input files were the output files of a lower level model, that is, the contrast between 2-back vs. 0-back. So in this higher level model there are only two factors, zygosity and performance, with two levels each.

I think the model described in part 1 of my original email fit with the principle of the file you sent me. Unfortunately the results seem unreasonable, because:
“… All the simple effect results (obtain from analyses in 1.1) of “high vs. low” looked exactly the same as the main effect of 2-back vs. 0-back (obtained from another one sample t test at the higher level). We thought this could not be the truth...”
Please refer to the first email for more details.

When defining the eb file for palm, we refereed your example for the HCP data as described in the paper titled " Multi-level block permutation". I attached our -eb file for your information (the first row here is only for labeling and was excluded in the analysis).


So my question remains. Did we do the model correctly (as described in part 1 of the first email; also see the testing & discussion in part 2 and 3)?  If not, how to do it please?


Thank you so much,
Kai