Thanks very much for your detailed response and explanations ! Yes, I constantly observed each cow at each observation over the same specified time frame, so I am going to use lc_hs1 instead of c_hs1 as offset variable. Hope it works for all the behaviours (response variables) I observed. Your example has clearly illustrated what's the difference between WEIGHT and OFFSET in my case. Thank you so much again ! ... View more

Hm, I apologize for not giving a code, thank you (initially I just wanted to know about the WEIGHT option as opposed to the OFFSET option) ! but no, I don't have a binary variable, just COUNTS of specific behaviours, which I observed multiple times/day (e.g. hourly I count a specific behaviour for a certain time (e.g. for 30 sec ....common methodology in ethology); for feasibility reasons it was not always possible and therefore number of times was not equal across all subjects. Yes, I sum across multiple observations to get my response variable for the model. (sumhs1 for example)., and I thought about taking the means across multiple observations. and yes, I used the number of times on which the response variable was based (c_hs1...number of times I counted hs during morning hours(=1) ) as offset term: model sumhs1= origin period mean_bg1 period*origin / offset= c_hs1 dist=poisson link=log solution; random farm farm*origin animal(farm*origin) period*farm period*animal(farm*origin); run; So, you mean offset=log(c_hs1) ? I am not sure if I understand your information correctly, that T measures the total "opportunity" for observing the behaviour; actually, it IS the total number of times e.g. some subjects could be observed every hour in the morning (between 8 and 12am; c_hs1=4, some could only be observed between 8 and 9 am, 9 and 10 am, but not between 10 and 11 am ...(c_hs1=2). ... View more

ok, no matter if means or totals (for the counts), I think I have to account for the unbalanced number of times (of observations) per subject - and I don't know if the WEIGHT option can do this. Haven't found an example. If someone knows, I would be pleased to get advice. ... View more

yes, I have counted the behaviours of the subjects repeatedly throughout the day: so MULTIPLE measurements per subject (no interest in temporal aspect), and created totals within certain hours of the day (morning, afternoon...) and use separate models for each time of the day, as I don't want to create more levels than I already have (farm--->subject(farm)---->(3)periods (of three days)). Because data are counts, I was advised to create totals so that there are still integers. Initially I also had means; actually I don't know what is more appropriate. Using means, I still have different number of times contributing to the means. In the end, I think I have to account for the number of times differing between the subjects anyway and I thought to do this using either the WEIGHT-option (?) or the OFFSET-option. My subjects are located on different farms, and I use a split-splot model (with farm and subject(within farm) as random effects; treatment and period as fixed effects; days within period = residual variation). The thing that I have no convergence using the OFFSET option worries me. HTH to understand my problem. Thanks very much for any help - Ka ... View more

Hi Peter, thanks very much for your hints. there's one mistake: I think I NEED an option which allows me to account for the different number of times I sampled my subjects (response variable = counts (counting a specific behaviour), which I summed up over the different number of times I counted the subjects), and I thought the offset-option would be the solution for it. I used the negbin distribution. Unfortunately, the model doesn't converge, if I do not omit the offset option for the different number of times. I think I HAVE TO account for the different number of times I counted the behaviour of the subjects, and wonder if there's another possibility using the WEIGHT option ? but I can't find an example explaining the use of this option...:-( There are just replicates (number of times counting the behaviours of the subjects), no repeated measures in my data. Any experience how to account for the unbalanced situation (different number of times counting the subjects) other than by using the OFFSET-option ? I would greatly appreciate your advice. Ka ... View more

Hi ! I have given it a try with glimmix using the option. unfortunately, there's no conversion, only if omit . Obviously, this is not the correct way when I have different numbers of times that I have observed/sampled my subjects, is it ? May anyone give me some advice how to correctly fit these count data ? Thank you very much in advance, Ka ... View more

The covariance structure type=sp(pow) time for my glimmix model (repeated date*period/subj=animal(farm) and the subject effect additional to the random farm effect in the random statement) seems to work fine (smaller -2 Res LogL and AIC ....closer to zero, respectively, than with cs), the same for type=UN@AR(1) for my mixed model; but it seems to me that my repeated structure is not correctly modeled as the dimensions table in my output(glimmix) tells me: Subjects = 1, and max Obs per subject = 234 while I have 26 subjects and 9 observations (3 repetitions per period) per subject and the model information table states that variance matrix = not blocked; Dimensions of G and R-side effects seem to be ok. similarly, in my output for the mixed model with the residual variance method = Parameter in the model information and subject=1 , max obs per subject = 234 in the dimensions table; So what's wrong with my model? proc glimmix: proc glimmix data=xxxx; class genotype farm animal period date; model x =genotype period genotype*period mean_bg / dist=negbin solution; random farm animal(farm); random date(period) / subject=animal(farm) type=sp(pow)(time) residual; run; I use the univariate format (one row per experimental unit at each timepoint) as emphasized in Littell et al, J. Anim. Sci. 1998. 76:1216–1231 for proc mixed/repeated measurements and other sources e.g. in http://www.sportsci.org/resource/stats/threetrials.html I assume it's the same for proc glimmix. In the SAS glimmix documentation (keyword: processing by subjects) I found the statement that "if a random statement does not have a subject=effect (as I do have in my model: random = farm), processing by subjects is not possible unless the random effect is a pure R-side overdispersion effect". As far as I understand, processing by subject is a question of how glimmix mathematically processes data but not a question of correctly modeling covariance structures, right ? if not, do I have to choose a subject for farm ??? Anyone a clou about this ? I also want to look at the significance of my random farm effect by omitting it from the model (in mixed as well as in glimmix) above, but SAS doesn't give a result in the output for the Likelihood Ratio Test. Is the Likelihood Ratio Test not possible with other effects in the random statement (e.g. the subject effect required by UN@AR(1) and sp(pow)(time)) ? Many thanks, Karin ... View more

My continuous covariate is a measure of ambient environmental conditions while I repeatedly (days) observed my experimental units (animals). To get estimates for the means of factor A (genotype) at the levels of B (periods) there's nothing else for it but fill the mean (across the repeated measurements) of mean_bg3, isn't it? Cheers, Karin ... View more

hm, what SteveDenham said was my assumption, but I was hoping that it should be possible to generate e.g. a line chart which shows immediately that the differences between lsmeans across levels of A at B=1 is not different from differences between lsmeans across levels of A at B=2 ...from my model without A*B in the model. mean_bg3 is a continuous covariate, I don't want to assign any values for the covariate mean_bg3 to be used for computing the lsmeans of A and B. Is this necessary? Does the proc means code you proposed for specifying the mean_bg3 value not produce the same as the option at means? Cheers, Ka ... View more

I skipped the thing about animal*farm inclusion in the random statement when using covariance structures as sp(pow)(time) and UN@AR(1) for the repeated structure. Does it imply animals to be a random sample rather than a selected sample or should between animal variability be modeled in this way regardless of such design features/premises? Cheers, Karin ... View more

Just a short post in reply (comments between your lines): > > First, nesting in SAS. You need to look at the > documentation for MIXED or GLM that covers this. The > parameterization of A nested in B and A crossed with > B is the same. That is why it makes no difference in > the rcorr. Here is what the online manual says: > Nested Effects > Nested effects are generated in the same manner as > crossed effects. Hence, the design columns generated > by the following two statements are the same (but the > ordering of the columns is different): > > model Y=A B(A); > > model Y=A A*B; > > The nesting operator in PROC MIXED is more a > notational convenience than an operation distinct > from crossing. Nested effects are typically > characterized by the property that the nested > variables never appear as main effects. Hm, meaning must not appear as main effect ? In my example the nested variable (within parentheses) period would appear in the model as main effect, since I am interested in the period effect. Did I miss the full meaning of this? > Note that nested effects are often distinguished from > interaction effects by the implied randomization > structure of the design. That is, they usually > indicate random effects within a fixed-effects > framework. The fact that random effects can be > modeled directly in the RANDOM statement might make > the specification of nested effects in the MODEL > statement unnecessary. Hm, I don't have any nested effect in the MODEL statement anyway, neither in the random statement, but in the repeated statement. > > The other part that I want to address is the doubly > repeated measures. This assumes that the correlation > of residuals amongst days is the same, no matter > which period we have. I was thinking that > observations on the close time points would be more > highly related, and probably similar, so that an > ar(1) or cs structure might fit. The separated time > points (i.e., periods) would have a different > relationship. > > If you want to deal with all at once without those > assumptions, you have unequal spacing and the spatial > power method goes after that with the assumption that > the errors are "constantly correlated" as in ar(1), > but that the points used to estimate that correlation > are not equally spaced. > I am certainly willing to try your recommendations. In Proc MIXED I'd prefer the former, and, if I get your point, to do that I have to model doubly repeated measures expressed as: repeated period day / subj=animal*farm type=UN@AR(1) with period in the first place instead of a nested repeated , right? does the order of these variables make any difference for the type=cov structure option? and in Proc GLIMMIX (for count data) maybe a spatial power method with sp(pow)(time) with repeated day*period / subj=animal*farm as the UN@AR(1) is not available. Would you confirm or reject my statements? Thanks a lot for your patient comments anyway, at least some of the points may be a bit clearer now. Karin ... View more