About SteveDenham

SteveDenham · ‎01-04-2012

The first thing that I would think of would be incomplete data, so that all degrees of freedom are used. I would guess that there is only one observation per site within wafer within lot--hence no residual variability and no variance due to site. What happens if you fit: proc nested; class lot wafer; var rho; run; The residual error in this case will be due to site within wafer within lot. Steve Denham

SteveDenham · ‎01-04-2012

There is an interesting work-around. Append the new data set to the old, but replace all of the response variable values in the new data set with missings " . " before appending. The fit will use only the original data, but the output data set (you'll need an OUTPUT statement that looks something like: OUTPUT OUT=predset p lcl ucl; This should give predicted values for all records with complete data for the independent variables, as well as confidence bounds. Steve Denham Message was edited by: Steve Denham

SteveDenham · ‎01-04-2012

The warning "Stopped because of infinite likelihood" raises some questions. Are there other statements in your PROC MIXED block (especially a REPEATED statement)? I'm also a bit baffled by the notes regarding the datasets. Surely these are not the datasets being passed to PROC MIXED, as for TEST there are only two variables, and for CONV, while there are enough variables for the specified model, there is only one record. Finally, it does look like PROC MIXED tries (cpu time almost 6 seconds) before the infinite likelihood problem is reached. That really does implicate the sweep operator. Have you tried using PROC GLIMMIX and using different optimization methods? Steve Denham

SteveDenham · ‎01-03-2012

You need an NLOPTIONS statement. Something like NLOPTIONS MAXITER=1000; should give you sufficient iterative ability to fit this. Also, if you are running under SAS v 9.3, PROC AUTOREG has an experimental CLASS statement, which might make sense to try and might make the interpretation a little easier, given the number of fish types you are looking at Steve Denham

SteveDenham · ‎01-03-2012

Probably a sweep operator "feature." Column order in the design matrix can often lead to near singularity when the sweep operator is applied during matrix inversion. What sort of notes in the log, or notations in the output listing were given when the model did not converge? Steve Denham

SteveDenham · ‎01-03-2012

What is the variable "diag" in the repeated statement? Since it is not included in the class statement, I must assume that it is a continuous variable that is unique for each record in the data set blood_pressure. In order for this to be valid, the dataset must be sorted on "diag." I also wonder that your input dataset and your G matrix dataset are identically named. I suspect that this will lead to problems, as the gdata= option requires somewhat different formatting. For further assistance, it would help if you posted any error or warning messages from the log. Steve Denham

SteveDenham · ‎12-05-2011

Take a look at PROC MULTTEST, especially at the example adjusting p values. Steve Denham

SteveDenham · ‎12-05-2011

I'm a bit confused here, because I come from a tradition that would consider sex as a fixed effect, and time and time-squared as covariates. The LSMEANS statement that you present would compare males to females at the average time AND at the average time-squared, and it may be the case that average time-squared does not equal (average time)*(average time), due to imbalance in data. Check the use of the AT= option in the LSMEANS statement. This may help clarify your thoughts. And don't think for a moment that you cannot include the interaction--it would simply mean differing slopes for the sexes. Steve Denham

SteveDenham · ‎11-21-2011

The best I can offer is to look at the on-line documentation for each, especially the overview of the procedures. My quick and dirty look: UCM is better for equally spaced time series analyses with trends, seasons and cycles. MIXED is better for unequally spaced data, where you may have some idea of the covariance structure of the observations, or where there are random effects that you wish to model. Steve Denham

SteveDenham · ‎11-17-2011

Quite possibly a mixture problem. It looks like 0.13% of the data significantly elevate the intercept. That can happen. Consider the mean net worth of a very rural county, where 200 people live, say $50,000 per person. All of a sudden Bill Gates moves in, worth say $20B. The mean net worth is now 1991 times as large, with only a change of 0.5% of the data. I would wager that those 1307 exceptional cases, when examined separately, tell you something quite interesting. Steve Denham

SteveDenham · ‎11-17-2011

The assumption of normality is greatly over-rated/misunderstood for most linear models, including the paired t-test. So long as the residuals, not the data themselves, are unimodal and relatively symmetric, the asymptotic p values should be a good guidepost. For most blood pressure data, this will be the case. But suppose you are testing the effect of a drug on lowering BP, and further suppose that the population is a mix of normotensive and hypertensives (which could be further divided by the degree of responsiveness to the drug). Now, you are probably in the realm of non-normality, and other approaches may be needed. But for this, it is a classic textbook case for application of a paired t-test. Steve Denham

SteveDenham · ‎11-11-2011

Wouldn't the REML estimates be adequate i.e., default estimates without the nobound option? Steve Denham

SteveDenham · ‎11-03-2011

That seems odd--I ran that code on existing data and got all the quantiles for each variable, named as per x1_p30 (30th percentile of x1), x1_p60 (60th percentile of x1), etc. Unfortunately, I agree that the decimal place thing doesn't work for PROC MEANS, and for PROC UNIVARIATE, it gets complicated with the pctlpts and pctlpre options. However, I got the following to work quite well, and the variables px12_5, px197_5, px22_5, px297_5, px32_5, px397_5, px42_5, and px497_5 are in the output dataset with labels that look like "the 2.5000 percentile, x1", etc.: proc univariate data=tryit; var x1-x4; output out=checkit2 pctlpts= 2.5 97.5 pctlpre= px1-px4; run; If you are only getting a single value, I have no idea what might be going on, unless you have BY, CLASS or WHERE statements that essentially partition the data before processing. Can you show your code? Steve Denham

SteveDenham · ‎11-03-2011

What about shifting to PROC MEANS from UNIVARIATE? This should give an output dataset named work.want with variables x1_P30, x1_P60, x1_P90, x2_P30, etc. proc means noprint data=have; var x1-x200; output out=want p30= p60= p90=/autoname; run; Steve Denham

SteveDenham · ‎11-03-2011

Var(1) and Var(2) are estimates of the residual error for the two ResponseIDs/dist variables, and Corr(2,1) the estimated correlation between the two. It is not surprising that the values change when the independent variable changes, as the relationship (solution values) will change, and thus the residual error would be different. The approach looks great. Steve Denham

Online Status	Offline
Date Last Visited	‎03-19-2026 03:00 PM

Re: Assessing Variable Redundancy for Mixed Effects Modeling

Re: Assessing Variable Redundancy for Mixed Effects Modeling

Re: Randomized block design and meaning of LSMEAN/STDERR

Re: Help with Restricted Cubic Splines : Code Optimization and Graphic...

Re: PROC POWER for Cox regression

Re: Repeated measures model executes in MIXED but not in GLIMMIX

Re: Model heteroscedasticity directly or use log transformation

Re: Model heteroscedasticity directly or use log transformation

Re: Repeated measures model executes in MIXED but not in GLIMMIX

Re: Repeated measures model executes in MIXED but not in GLIMMIX

Re: Passing TESTVALUE option in LSMESTIMATE statement in glimmix

Re: What is the "estimate" in the SolutionR output of the proc mixed. ...

Re: Assessing Variable Redundancy for Mixed Effects Modeling

Re: SAS OnDemand Outpm option in Proc Mixed

Re: question about lsmean pdiff=;option in proc glm step, st102d03, SA...

Re: Proc mixed, defining data structure for desired comparison (Random...

Re: Assessing Variable Redundancy for Mixed Effects Modeling

Re: Assessing Variable Redundancy for Mixed Effects Modeling

Re: Help with Restricted Cubic Splines : Code Optimization and Graphic...

Re: Repeated measures model executes in MIXED but not in GLIMMIX

SAS Nested Procedure shows dots

Re: How to score (with 95% confidence values) each observation in GLM ...

Order of variables in Proc Mixed

Proc Autoreg with hetero statement, convergence error

Order of variables in Proc Mixed

using proc mixed for meta-analysis

proc surveyreg and post-hoc comparisons

Longitudinal mixed models - lsmeans for a fixed effect with continuous...

proc mixed or proc ucm?

Overall mean of general linear model

Trying to determine correct statistical test to use - need assistance

Bayesian estimate of the within and between variance with PROC MIXED ?

Re: Quantiles and Sum of Variables...

Re: Quantiles and Sum of Variables...

Using GLIMMIX for 2 binary dependent variables

SAS Analytics Explorers