Someone in the Community might be able to help if you provided more information about your study, and if you showed what you have tried so far (in other words, your MIXED code).   For example, What is your subject, on which you took repeated measurements? How many BREEDs? How many subjects in each BREED? What is TIME, how many levels does it have, are you using TIME as a classification or numeric variable? What is WEEK, how many levels does it have, are you using WEEK as a classification or numeric variable? What is the relationship between TIME and WEEK?   I hope this helps.   ... View more

Also, your MODEL statement does not follow the hierarchical principle for interaction:    When interactions are included in the model, the main effects needs to be included in the model as well. The main effects needs to be included even if the individual variables are not significant in the model. https://datascientia.blog/2017/08/27/dss-p9-interactions/    For example, if you include "week*hrs2_group*operator" then you also include the 3 main effects and the 3 two-way interactions.   This rule should be followed most of the time; if you don't follow it, then you should know why you aren't.   ... View more

As I think you are saying, algebraically, the difference   exp(B0 +B1) - exp(B0+B2) as computed by ESTIMATE   is not equal to exp( (B0+B1) - (B0+B2) ) = exp( B1 - B2 ) as computed by ESTIMATE   I'd suggest using ESTIMATE to compute (B1 - B2) and whatever interval you want (e.g., a CI or +/- SE), and then back-transforming the difference and the interval endpoints.   EDIT: Oops, my suggestion is wrong, and I've striked it out. My apologies. You could compute the point estimate for exp(B0+B1) - exp(B0+B2) by hand using output from the first two ESTIMATE statements.    I am not convinced that the ESTIMATE statements are correctly specified. The LSMESTIMATE statement is more foolproof than the ESTIMATE statement, and you might check into that. See CONTRAST and ESTIMATE Statements Made Easy: The LSMESTIMATE Statement. You can check your specifications using output from LSMEANS. ... View more

Your description does not give us enough information to determine whether the statistical model is correct. For example, how many levels of CASE are there, and how does CASE relate to TRIAL? Are there 3 CASEs with one TRIAL each? Or 3 TRIALs for each CASE? What research question do you have that would be addressed by 12 SCANs in each TRIAL?    Your residual plot and data plot show that there is an upper bound (which is zero) to your response "PlumbResult_0504_LateralDeviatio". Neither the lognormal nor the gamma distribution is appropriate for data with an upper bound; both the lognormal and the gamma have a lower bound at zero and an upper bound of infinity. Both should have failed miserably with a response with negative values (the log of zero is not defined, and I would guess that there was a message to that effect in the log window; always pay attention to the log window).   So, we need to know more about what your response is measuring, in addition to more about your experimental design. Guessing wildly, you might have more luck redefining your response as (-1)*response, if that was sensible in context; that redefined response might follow the exponential distribution, and then the gamma might work (although gamma mixed models can be very persnickety).    Given your current description, I doubt your RANDOM specifications are right but we await more information.   ... View more

You can obtain comparisons of simple effects using the SLICE statement that is available in several procedures, including MIXED and GLIMMIX but not GLM, see the SLICE Statement documentation   or you can use the SLICE option on the LSMEANS or LSMESTIMATE statement in the MIXED and GLIMMIX procedures.   I hope this helps.       ... View more

This example Example 38.15 Comparing Multiple B-Splines appears to do what you want.   ... View more

By "interaction with PRODUCT*V1" do you mean the effect of PRODUCT when VISIT=V1? If so, then this comparison is known as a "simple effect". It is not interaction.   An interaction would be implied when the effect of PRODUCT when VISIT=V1 is different than the effect of PRODUCT when VISIT=V2. Perhaps this is what you mean by "global interaction".   "Meaningful" is not the same thing as "significant". The effect can be meaningful (in context) without enough evidence for (statistical) significance if a p-value is large-ish (which some people might interpret as p > 0.05) but sample size is small-ish (implying a potential lack of adequate power). The effect can be (statistically) significant without being meaningful (in context) if the sample size is large-ish and the p-value is small-ish.    It is possible for the "global" interaction test to have a large-ish p-value while one of the simple effects components has a p-value that is small-ish. My interpretation in that scenario generally is that there is too much uncertainty to conclude that the interaction is "significant". But both "meaningful" and "significant" are context-specific, and there are many nuances.    The distinction between meaningful and significant, and the role of a p-value in statistical decisions, has an extensive and argumentative literature. The recent special issue of The American Statistician https://www.tandfonline.com/doi/full/10.1080/00031305.2019.1583913 is a good place to start.   I hope this helps.     ... View more

The coefficients are for counts on the link (here, log) scale. ... View more