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
  • 2,243 Likes Given
  • 255 Solutions
  • 2,422 Likes Received
About
Linear models

Re: Using estimate statement in proc surveyreg

by Jade | Level 19 in Statistical Procedures
‎10-20-2020 09:03 AM
1 Like
‎10-20-2020 09:03 AM
1 Like
I think what you are calling estimates are estimates of the parameters, which under the non-full rank parameterization used, sets the last category to zero.  Using the LSMEANS statement combines the necessary elements to calculate the expected population mean for each level of a categorical variable.   This is a case where using the LSMEANS and LSMESTIMATE statement is an advantage;  Try adding this to your code:   lsmeans vaper|riagender/e; lsmestimate riagender "Males vs Females, averaged over vaper status' -1 1/exp; lsmestimate vaper*riagender 'Males vs Females, vaper status=no' -1 1 0 0, 'Males vs Females, vaper status=yes' 0 0 -1 1/exp; You may need to rearrange the coefficients in the second LSMESTIMATE statement to reflect the parameterization given in the LSMEANS statement by the /e option.   SteveDenham     ... View more

Re: MCMC Lognormal.

by Jade | Level 19 in Statistical Procedures
‎10-19-2020 08:22 AM
1 Like
‎10-19-2020 08:22 AM
1 Like
The primary reason is that exponentiating the mean of a log normal yields the geometric mean, which is almost certainly less than the mean assuming normality.  Take a look at any graphical description of the lognormal distribution, or see this https://en.wikipedia.org/wiki/Log-normal_distribution .  There you will see that the mean (expected value) of the log-normal is exp(mu + variance/2).  The geometric mean is an approximate estimator of the median, which in a right skewed distribution is less than the mean.   SteveDenham ... View more

Re: Checking model assumptions in Proc GLIMMIX

by Jade | Level 19 in Statistical Procedures
‎10-19-2020 08:09 AM
‎10-19-2020 08:09 AM
This is how I usually write an R side covariance:   proc glimmix data=BAL plots=all; class number forage week; model freq_PMN/TNCC = forage week week*forage/solution link=logit; random week/subject=number type=ar(1) residual; run; You might consider an AR+RE model as well:   proc glimmix data=BAL plots=all; class number forage week; model freq_PMN/TNCC = forage week week*forage/solution link=logit; random intercept/subject=number; random week/subject=number type=ar(1) residual; run; Good luck with the reviewer.   SteveDenham     ... View more

Re: Checking model assumptions in Proc GLIMMIX

by Jade | Level 19 in Statistical Procedures
‎10-16-2020 08:31 AM
2 Likes
‎10-16-2020 08:31 AM
2 Likes
Ordinarily, I would include week in the random residual statement, and find this approach interesting.   But that doesn't answer the question.  So... Since this is modeling R side covariances, and there is a logit link, I don't see how you can satisfy the reviewer.  In the logit space, the residual should be approximately normal, so maybe you can extract the residuals and do a QQ plot. Except that these aren't really residuals, but are BLUPs.  You could also output variances, although what to do with them is an issue  Testing for normality could also be done, but that requires that the residuals be independent, and these may be clustered by subject. A nonparametric (in the distribution-free sense) test might work (KS? Cramer-von MIses?).  Stroup's text doesn't go into this issue, so I am winging it here. I trust QQ plots more than I trust normality tests, so that would be my choice. It is just getting the right things to plot.   SteveDenham ... View more

Re: Proc Power without Reference MEANDIFF

by Jade | Level 19 in Statistical Procedures
‎10-15-2020 02:36 PM
‎10-15-2020 02:36 PM
Those would be the K values, with the pooled standard deviation set to 1. I am trying to get this on the basis of a normal(0,1) distribution.  Now with unequal standard deviations, this would be trickier.  First, you must change to TEST=DIFF_SATT to get unequal variances per group.  Then, in the output, you'll see for each value of MEANDIFF a value called actual alpha and a power.  The actual alpha is greater than the nominal 0.05 because Satterthwaite's approximation introduces a bias.  In this case the mean difference is NOT in terms of one or the other input groupstddev values, but follows the approximation found in the Details section of the PROC POWER documentation.   standard error of the difference = sqrt (sone * sone/none  + stwo*stwo/ntwo), where sone is the standard deviation of group one, stwo is the standard deviation of group two, none is the number in group one, and ntwo is the number in group two.  For your case, sone=1,none=24, stwo=2, ntwo=8, and standard error of the difference is sqrt(13/24)=0.74.  So in this case, the mean differences are 0.2*standard error of the difference up to 2.4*standard error of the difference.  You could plug in non-scaled values, if you have them.   SteveDenham ... View more

Re: Proc Power without Reference MEANDIFF

by Jade | Level 19 in Statistical Procedures
‎10-15-2020 12:45 PM
‎10-15-2020 12:45 PM
Yes.  In fact, my mean diff values can be viewed as K * stddev, if you set stddev=1.  The whole idea of doing that is that no matter what the exact value of the standard deviation is, you can calculate the power if the difference to be detected (the meandiff value) is scaled. I hope that makes sense - it took me a while to learn that shortcut.  It really comes in handy when you don't have pilot data.   SteveDenham ... View more

Re: PROC MIXED backtransforming LSMEANS Diffs or LSMESTIMATE compariso...

by Jade | Level 19 in Statistical Procedures
‎10-15-2020 11:04 AM
1 Like
‎10-15-2020 11:04 AM
1 Like
Agreed - the p values obtained are based on differences in the log normal space.  My thought was to be able to present the results on the original scale as well.   SteveDenham ... View more

Re: Proc Power without Reference MEANDIFF

by Jade | Level 19 in Statistical Procedures
‎10-15-2020 08:48 AM
‎10-15-2020 08:48 AM
Try running with a variety of MEANDIFF values, and examining the power estimates. If you look at acceptable power levels (say 0.8), then you could see what the detectable difference associated with that power level.  Try this code:   proc power; twosamplemeans test=diff groupstddevs = 1| 2 ntotal = 32 meandiff = 0.2 to 2.4 by 0.2 power = . groupweights=(3,1); ods output output=output; run; SteveDenham   ... View more

Re: PROC MIXED backtransforming LSMEANS Diffs or LSMESTIMATE compariso...

by Jade | Level 19 in Statistical Procedures
‎10-15-2020 08:37 AM
1 Like
‎10-15-2020 08:37 AM
1 Like
Consider shifting to GLIMMIX, and using either a log link or a lognormal distribution.  For the former, the ILINK option will work for the differences between time points.  For the latter (log normal), you can derive the significances for differences in the log space (which is what you are getting from a log transform in MIXED).  Actual differences will have to be computed externally in a DATA step.  You can either look at differences between geometric means (simple exponentiating of the lsmean estimates), or at differences in the expected values (complex exponentiation that takes 3 lines of code, followed by the difference).  Note that neither of these backtransforms yields the standard error of the difference, although the second method depends on calculating a variance, from which a standard error of the backtransformed mean can be obtained.  See the documentation for the DIST= option in the MODEL statement of PROC GLIMMIX for the equations needed.   SteveDenham ... View more

Re: Use both fixed effect dummies and newey west t stat

by Jade | Level 19 in SAS Forecasting and Econometrics
‎10-14-2020 08:00 AM
‎10-14-2020 08:00 AM
Moved to Forecasting and Econometrics, in hope of attracting answers from those familiar with SAS/ETS.   SteveDenham ... View more

Re: SEs for fixed effects affected by solution in random statement in ...

by Jade | Level 19 in Statistical Procedures
‎10-13-2020 09:58 AM
‎10-13-2020 09:58 AM
Good question.  I have a hunch you will need to take this up with Tech Support.  When you get an answer, please post it here.  This is the kind of unexpected behavior that some of us need to consider.   SteveDenham ... View more

Re: LSMEANS in time series

by Jade | Level 19 in Statistical Procedures
‎10-13-2020 09:50 AM
‎10-13-2020 09:50 AM
Question 1: What is 'spline' doing here? Question 2: Isn't 'intervention' completely confounded with Block (i.e., Intervention=pre when Block<24, intervention=post when Block >=24?   Perhaps you could fit the segmented model without 'intervention' in the model.  Then when you get to the GLM phase, use an ESTIMATE or LSMESTIMATE to get the marginal means for intervention pre and intervention post.  I still don't know how to get around the spline issue - perhaps an AT option could be used in the LSMESTIMATE statement.   SteveDenham ... View more

Re: Creating a reference group and using it in a logistic regression m...

by Jade | Level 19 in Statistical Procedures
‎10-13-2020 09:32 AM
1 Like
‎10-13-2020 09:32 AM
1 Like
It runs without errors (but with the quasi-separation warning) on my machine.  Since you are devising a reference group based on sum(a1-a4)=0, you may wish to try one of the following: 1. Remove a1 thru a4 as predictors This would give the OR for the levels of refgroup. 2. Use a BY refgroup statement and remove refgroup as a predictor.  This would give the ORs for each refgroup separately.    I don't think you are going to be able to get accurate estimates if both refgroup and a1 thru a4 are included, primarily due to multicollinearity.   SteveDenham ... View more

Re: proc mix: Reverse the direction of difference using PDIFF

by Jade | Level 19 in Statistical Procedures
‎10-13-2020 09:03 AM
‎10-13-2020 09:03 AM
@FreelanceReinh , nice to see that your answer matches mine.  This thread is duplicated, and I am not sure how to merge the two.   SteveDenham ... View more

Re: How to reverse the direction of difference using pdiff option in p...

by Jade | Level 19 in Statistical Procedures
‎10-13-2020 08:57 AM
1 Like
‎10-13-2020 08:57 AM
1 Like
Use an LSMESTIMATE statement that utilizes the coefficients in the direction of interest.  For example,   lsmestimate treatn 'Placebo - trt 1' -1 0 1,                   'Placebo - trt 2' 0 -1 1;          SteveDenham ... 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
View All
Quick links