cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Search instead for 
Did you mean: 
SteveDenham
‎03-19-2026
  • 3,978 Posts
  • 2,240 Likes Given
  • 255 Solutions
  • 2,422 Likes Received
About
Linear models

Covtest in Proc Glimmex

by Jade | Level 19 in Statistical Procedures
‎02-20-2012 01:36 PM
3 Likes
‎02-20-2012 01:36 PM
3 Likes
Since I am running on 9.3, I'm not sure which options were viable in 9.2 and 9.22 (I am pretty sure that there was an update).  The code is 'red' in 9.3 as well. It looks like all of the options are available in 9.22, so it should be working. /turn on grasping at straws mode So, just how big is your G matrix, and how many obs do you have?  Maybe it is an estimation problem. This one might need a visit to tech services. Steve Denham ... View more

Covtest in Proc Glimmex

by Jade | Level 19 in Statistical Procedures
‎02-20-2012 08:03 AM
3 Likes
‎02-20-2012 08:03 AM
3 Likes
It is possible that there are convergence problems of some sort--I have run into this with covtest homogeneity. Try adding a restart option, e.g. covtest 'G reduced' zerog/restart; I can't guarantee this will help, but it would be the first thing I would try in this case.  Number two would be to change the test to that for a diagonal G matrix.  With the full parameterization under the Cholesky structure, this may at least run. Good luck. Steve Denham ... View more

Offsetting in proc genmod

by Jade | Level 19 in Statistical Procedures
‎02-20-2012 07:52 AM
‎02-20-2012 07:52 AM
The offset option essentially scales the response variable.  I think you are asking about scaling independent variables.  Is that correct?  If so, you will probably have to apply a dataset transformation, or include protgramming statements within PROC GENMOD. Steve Denham ... View more

Data for variability

by Jade | Level 19 in Statistical Procedures
‎02-17-2012 08:21 AM
‎02-17-2012 08:21 AM
The first line tests whether the mean of group A is different from zero, the second whether the mean of group B is different from zero, and the third whether the difference between A and B is itself different from zero.  Beware of subtracting between confidence bounds--you could have a significant difference without the difference being greater than zero.  And it scares me a little that you have multiple groups and are using multiple t-tests to compare them two at a time.  That sort of comparison is handled a lot better by analysis of variance techniques. Steve Denham ... View more

why selection=stepwise is not one very good way to do regression?

by Jade | Level 19 in Statistical Procedures
‎02-17-2012 08:16 AM
1 Like
‎02-17-2012 08:16 AM
1 Like
Regarding the last statement: Stepwise regression would still yield biased results.  It is a matter of sampling from the population. You cannot get around it. Now what you could do, given the abilities specified, is measure all possible variables on every individual in the population, and fit that by regression.  And watch collinearity kill the interpretation. In my opinion, and I stress that this is only an opinion, regression is just not quite the right tool for data exploration.  It is a great tool for finding the degree of relationship for pre-specified variables. In these days of big data, and in the days to come of even bigger data, I wonder if the whole branch of statistics that falls under "linear models" like regression, ANOVA, GLMMs, etc. will be considered the equivalent of steam power. Steve Denham ... View more

linear regression underestimation

by Jade | Level 19 in Statistical Procedures
‎02-15-2012 08:01 AM
‎02-15-2012 08:01 AM
Thanks, Doc.  I feel like I may have committed a Type 3 error--answered the wrong question correctly.  The underestimation of the mean by using a log transform followed by a simple exponentiation backtransform is rampant in the biological field as well.  Getting a proper estimate of the standard error of predicted values (LSmeans) is even worse.  At least the documentation for PROC GLIMMIX discusses what kind of backtransform is needed for lognormal distributions.  Still, the geometric mean or median estimate that you get with a simple exponentiation may be more meaningful than the mean--after all, that is part of why the transformation is applied: the data are skewed and the expectation of the variable is "inflated". Steve Denham ... View more

linear regression underestimation

by Jade | Level 19 in Statistical Procedures
‎02-14-2012 06:33 AM
‎02-14-2012 06:33 AM
Hard to say without more information.  However, in PROC TRANSREG you can transform both the right hand side and left hand side variables--specifying METHOD=MORALS or METHOD=REDUNDANCY may address the underestimation problem. Steve Denham ... View more

incomplete block design 2 or 3 factors

by Jade | Level 19 in Statistical Procedures
‎02-13-2012 02:07 PM
‎02-13-2012 02:07 PM
How about: proc mixed data=zzz; class factor1 factor2 block; model response=factor1 factor2 factor1*factor2/solution ddfm=kr; random block; lsmeans factor1 factor2 factor1*factor2/cl; run; The key, of course, is in preparing the dataset zzz, so that it correctly reflects the incomplete block design. Steve Denham ... View more

Correct coding of glimmix for both g and r-side random effects

by Jade | Level 19 in Statistical Procedures
‎02-13-2012 02:04 PM
1 Like
‎02-13-2012 02:04 PM
1 Like
Replace.  Otherwise, it will try to estimate the between land_types twice, and set one to zero, or some other odd thing. Final suggested code: proc glimmix data=XX ; class Year Land_Type LandID PCSite; model occ(event='1')=vht /dist=binary link=logit solution; random intercept Land_type/subject=LandID; random Year/Subject=PCSite type=chol residual; run; Hope this works... Steve Denham ... View more

Correct coding of glimmix for both g and r-side random effects

by Jade | Level 19 in Statistical Procedures
‎02-13-2012 01:48 PM
‎02-13-2012 01:48 PM
If I understand everything correctly, you have two observations for each PCSite--one in year1 and one in year2.  If this is the case, you might as well specify type=chol (an unstructured Cholesky parameterization) for the covariance structure.  There are only three R-side parameters to be estimated.  So I would only slightly change the R side specification to: random year/type=chol subject=pcsite residual; and make sure that pcsite is included in the class statement. Now back to the G side. I get the impression that landid is unique, and so a given sequence of landid's is nested within each landscape, as some sort of sub-sample.  How does landid differ from pcsite?  This will motivate what I say next.  Are the point count sites clustered within the landid's?  If so, then maybe something like: random intercept land_type/subject=landid; would be appropriate.  This should give two variance components: one arising from landid's within land_types and one arising from between land_types. If the pcsites are NOT clustered within the landid's, then we need to consider just what the experimental unit is, and maybe start to think about spatial location covariance structures. Steve Denham ... View more

Plotting the Nelson-Aalen estimate

by Jade | Level 19 in Statistical Procedures
‎02-13-2012 01:23 PM
‎02-13-2012 01:23 PM
And for a smoothed line rather than a step function: proc sgplot data=ple; pbspline x=dur y=cumhaz; run; Steve Denham ... View more

Plotting the Nelson-Aalen estimate

by Jade | Level 19 in Statistical Procedures
‎02-13-2012 08:05 AM
‎02-13-2012 08:05 AM
Use the ods output statement I gave (ods output productlimitestimates=ple) to get a dataset with the cumulative hazards estimates.  Then plot it using PROC SGPLOT. Steve Denham ... View more

Plotting the Nelson-Aalen estimate

by Jade | Level 19 in Statistical Procedures
‎02-10-2012 08:01 AM
‎02-10-2012 08:01 AM
Not real sure if this is what you want, but it did run and gave a smoothed plot of the estimated hazard rate. proc lifetest data=hepatitis nelson plots=h; time dur*status(0); ods output productlimitestimates=ple; run; The dataset ple has the Nelson-Aalen estimates (along with some other stuff).  You could then plot the cumulatives using PROC SGPLOT, using the PBSPLINE option.  I am afraid that is outside my skill set, though. Steve Denham ... View more

Zero Inflated Poisson using Proc Genmod

by Jade | Level 19 in Statistical Procedures
‎02-09-2012 07:58 AM
3 Likes
‎02-09-2012 07:58 AM
3 Likes
I don't have a good answer, but I know who does and how to find their answers: Head over to the archives for the SAS listserv at http://www.listserv.uga.edu/cgi-bin/wa?S1=sas-l and search on ZIP and NLMIXED.  You should find a ton of entries, many of which address your problem. Steve Denham ... View more

Means tables from ANCOVA in SurveyReg

by Jade | Level 19 in Statistical Procedures
‎02-09-2012 07:45 AM
‎02-09-2012 07:45 AM
Linda, Have you specified an LSMEANS statement? From what you have presented, I assume you have something like: proc surveyreg data=have; strata sex; class sex; model depvar=sex othervar sex*othervar/solution; weight weight; run; Add the LSMEANS statement that looks something like the following, and it should give you marginal means, adjusted for the covariate 'othervar'.  Note that these are predicted marginals for a hypothetical balanced population. proc surveyreg data=have; strata sex; class sex; model depvar=sex othervar sex*othervar/solution; weight weight; lsmeans sex; run; There are several options available to the LSMEANS statement that you might want to explore.  Also, I specified a model that had different slopes for each sex in the ANCOVA--you may wish to fit a common slopes model which would not include the interaction term. Good luck. Steve Denham ... View more
Contact Me
Online Status
Offline
Date Last Visited
‎03-19-2026 03:00 PM
Group Hubs for SteveDenham
View All
Likes given to
Quick links