Two-level, one-way contrast and ordering of levels

[myllym]

Two-level, one-way contrast and ordering of levels

[myllym]

Thank you for your comment and interest! At the moment, the contrasts are indeed hard-coded to the specific style, but I will think about the possibility to allow the user to set them.

I am sorry for the missing information about the order in the default title when there is only one plot/comparison. If your test output is saved in an R object res, you can find the title that you would like to have in attr(res, "labels").

[jmhrrsn]

Thank you for the quick response!

Just to confirm: am I correct in thinking that multiplying obs, central, hi and lo by -1 and then exchanging lo and hi will give me the result I would have gotten if the contrast were done in the other direction?

[myllym]

I realise it can be a bit confusing with the different contrasts. In graph.flm, first, to obtain the regression coefficients for each group, the constrasts option "contr.sum" is used for the lm function. If you use contrasts = FALSE , the coefficients are used as the test statistic as such. If you use constrasts = TRUE, then differences of the coefficients are used as the test statistic. Given the coefficients in a specific order that is obtained through the use of lm and dummy.coef, the differences are taken currently by checking only couples i and j where i < j and reducing j from i. E.g. groups 1, 2, 3 leads to comparisons 1-2, 1-3, 2-3. I hope this clarifies.

[jmhrrsn]

Thank you! For others' reference, the workaround I discovered based on your explanation was to relevel my factor with the treatment condition as the reference level.

[myllym]

Hi,
this took some time, but now the function graph.flm has a further option NULL for the argument contrasts. If contrasts = NULL the coefficients given by lm are used directly as the test statistic. (The contrasts of lm can be controlled by options.)
-Mari