Re: [R] Differences in output of lme() when introducing interactions
I clearly am going to have to improve my stats knowledge by reading McPhearson. To heck with Senn- too complicated. :) Thanks Terry. John Kane Kingston ON Canada -Original Message- From: thern...@mayo.edu Sent: Thu, 23 Jul 2015 14:07:00 -0500 To: r.tur...@auckland.ac.nz, thern...@mayo.edu Subject: Re: [R] Differences in output of lme() when introducing interactions The following are in parody (but like all good parody correct wrt the salient features). The musings of Guernsey McPhearson http://www.senns.demon.co.uk/wprose.html#Mixed http://www.senns.demon.co.uk/wprose.html#FDA In formal publication: Senn, Statistical Issues in Drug Development, second edition, Chapter 14: Multicentre Trials Senn, The many modes of meta, Drug information journal, 34:535-549, 2000. The second points out that in a meta analysis no one would ever consider giving both large and small trials equal weights, and relates that to several other bits of standard practice. The 'equal weights' notion embedded in a fixed effects model + SAS type 3 is an isolated backwater. Terry T. PS. The Devils' Drug Development Dictionary at the same source has some gems. Three rather random choices: Bayesian - One who, vaguely expecting a horse and catching a glimpse of a donkey, strongly concludes he has seen a mule. Medical Statistician - One who won't accept that Columbus discovered America because he said he was looking for India in the trial Plan. Trend Towards Significance - An ever present help in times of trouble. On 07/22/2015 06:02 PM, Rolf Turner wrote: On 23/07/15 01:15, Therneau, Terry M., Ph.D. wrote: SNIP 3. Should you ever use it [i.e. Type III SS]? No. There is a very strong inverse correlation between understand what it really is and recommend its use. Stephen Senn has written very intellgently on the issues. Terry --- can you please supply an explicit citation? Ta. cheers, Rolf __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. FREE ONLINE PHOTOSHARING - Share your photos online with your friends and family! Visit http://www.inbox.com/photosharing to find out more! __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Differences in output of lme() when introducing interactions
On 24/07/15 07:07, Therneau, Terry M., Ph.D. wrote: The following are in parody (but like all good parody correct wrt the salient features). The musings of Guernsey McPhearson http://www.senns.demon.co.uk/wprose.html#Mixed http://www.senns.demon.co.uk/wprose.html#FDA In formal publication: Senn, Statistical Issues in Drug Development, second edition, Chapter 14: Multicentre Trials Senn, The many modes of meta, Drug information journal, 34:535-549, 2000. The second points out that in a meta analysis no one would ever consider giving both large and small trials equal weights, and relates that to several other bits of standard practice. The 'equal weights' notion embedded in a fixed effects model + SAS type 3 is an isolated backwater. Terry T. PS. The Devils' Drug Development Dictionary at the same source has some gems. Three rather random choices: Bayesian - One who, vaguely expecting a horse and catching a glimpse of a donkey, strongly concludes he has seen a mule. Medical Statistician - One who won't accept that Columbus discovered America because he said he was looking for India in the trial Plan. Trend Towards Significance - An ever present help in times of trouble. On 07/22/2015 06:02 PM, Rolf Turner wrote: On 23/07/15 01:15, Therneau, Terry M., Ph.D. wrote: SNIP 3. Should you ever use it [i.e. Type III SS]? No. There is a very strong inverse correlation between understand what it really is and recommend its use. Stephen Senn has written very intellgently on the issues. Terry --- can you please supply an explicit citation? Ta. Thanks Terry! cheers, Rolf -- Technical Editor ANZJS Department of Statistics University of Auckland Phone: +64-9-373-7599 ext. 88276 __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Differences in output of lme() when introducing interactions
The following are in parody (but like all good parody correct wrt the salient features). The musings of Guernsey McPhearson http://www.senns.demon.co.uk/wprose.html#Mixed http://www.senns.demon.co.uk/wprose.html#FDA In formal publication: Senn, Statistical Issues in Drug Development, second edition, Chapter 14: Multicentre Trials Senn, The many modes of meta, Drug information journal, 34:535-549, 2000. The second points out that in a meta analysis no one would ever consider giving both large and small trials equal weights, and relates that to several other bits of standard practice. The 'equal weights' notion embedded in a fixed effects model + SAS type 3 is an isolated backwater. Terry T. PS. The Devils' Drug Development Dictionary at the same source has some gems. Three rather random choices: Bayesian - One who, vaguely expecting a horse and catching a glimpse of a donkey, strongly concludes he has seen a mule. Medical Statistician - One who won't accept that Columbus discovered America because he said he was looking for India in the trial Plan. Trend Towards Significance - An ever present help in times of trouble. On 07/22/2015 06:02 PM, Rolf Turner wrote: On 23/07/15 01:15, Therneau, Terry M., Ph.D. wrote: SNIP 3. Should you ever use it [i.e. Type III SS]? No. There is a very strong inverse correlation between understand what it really is and recommend its use. Stephen Senn has written very intellgently on the issues. Terry --- can you please supply an explicit citation? Ta. cheers, Rolf __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Differences in output of lme() when introducing interactions
Type III is a peculiarity of SAS, which has taken root in the world. There are 3 main questions wrt to it: 1. How to compute it (outside of SAS). There is a trick using contr.treatment coding that works if the design has no missing factor combinations, your post has a link to such a description. The SAS documentation is very obtuse, thus almost no one knows how to compute the general case. 2. What is it? It is a population average. The predicted average treatment effect in a balanced population-- one where all the factor combinations appeared the same number of times. One way to compute 'type 3' is to create such a data set, get all the predicted values, and then take the average prediction for treatment A, average for treatment B, average for C, ... and test are these averages the same. The algorithm of #1 above leads to another explanation which is a false trail, in my opinion. 3. Should you ever use it? No. There is a very strong inverse correlation between understand what it really is and recommend its use. Stephen Senn has written very intellgently on the issues. Terry Therneau On 07/22/2015 05:00 AM, r-help-requ...@r-project.org wrote: Dear Michael, thanks a lot. I am studying the marginality and I came across to this post: http://www.ats.ucla.edu/stat/r/faq/type3.htm Do you think that the procedure there described is the right one to solve my problem? Would you have any other online resources to suggest especially dealing with R? My department does not have a statician, so I have to find a solution with my own capacities. Thanks in advance Angelo __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Differences in output of lme() when introducing interactions
On 23/07/15 01:15, Therneau, Terry M., Ph.D. wrote: SNIP 3. Should you ever use it [i.e. Type III SS]? No. There is a very strong inverse correlation between understand what it really is and recommend its use. Stephen Senn has written very intellgently on the issues. Terry --- can you please supply an explicit citation? Ta. cheers, Rolf -- Technical Editor ANZJS Department of Statistics University of Auckland Phone: +64-9-373-7599 ext. 88276 __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Differences in output of lme() when introducing interactions
Hi, In addition to Terry’s great comments below, as this subject has come up frequently over the years, there is also a great document by Bill Venables that is valuable reading: Exegeses on Linear Models http://www.stats.ox.ac.uk/pub/MASS3/Exegeses.pdf Regards, Marc Schwartz On Jul 22, 2015, at 8:15 AM, Therneau, Terry M., Ph.D. thern...@mayo.edu wrote: Type III is a peculiarity of SAS, which has taken root in the world. There are 3 main questions wrt to it: 1. How to compute it (outside of SAS). There is a trick using contr.treatment coding that works if the design has no missing factor combinations, your post has a link to such a description. The SAS documentation is very obtuse, thus almost no one knows how to compute the general case. 2. What is it? It is a population average. The predicted average treatment effect in a balanced population-- one where all the factor combinations appeared the same number of times. One way to compute 'type 3' is to create such a data set, get all the predicted values, and then take the average prediction for treatment A, average for treatment B, average for C, ... and test are these averages the same. The algorithm of #1 above leads to another explanation which is a false trail, in my opinion. 3. Should you ever use it? No. There is a very strong inverse correlation between understand what it really is and recommend its use. Stephen Senn has written very intellgently on the issues. Terry Therneau On 07/22/2015 05:00 AM, r-help-requ...@r-project.org wrote: Dear Michael, thanks a lot. I am studying the marginality and I came across to this post: http://www.ats.ucla.edu/stat/r/faq/type3.htm Do you think that the procedure there described is the right one to solve my problem? Would you have any other online resources to suggest especially dealing with R? My department does not have a statician, so I have to find a solution with my own capacities. Thanks in advance Angelo __ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Differences in output of lme() when introducing interactions
In-line On 20/07/2015 15:10, angelo.arc...@virgilio.it wrote: Dear List Members, I am searching for correlations between a dependent variable and a factor or a combination of factors in a repeated measure design. So I use lme() function in R. However, I am getting very different results depending on whether I add on the lme formula various factors compared to when only one is present. If a factor is found to be significant, shouldn't remain significant also when more factors are introduced in the model? The short answer is 'No'. The long answer is contained in any good book on statistics which you really need to have by your side as the long answer is too long to include in an email. I give an example of the outputs I get using the two models. In the first model I use one single factor: library(nlme) summary(lme(Mode ~ Weight, data = Gravel_ds, random = ~1 | Subject)) Linear mixed-effects model fit by REML Data: Gravel_ds AIC BIC logLik 2119.28 2130.154 -1055.64 Random effects: Formula: ~1 | Subject (Intercept) Residual StdDev:1952.495 2496.424 Fixed effects: Mode ~ Weight Value Std.Error DF t-value p-value (Intercept) 10308.966 2319.0711 95 4.445299 0.000 Weight-99.036 32.3094 17 -3.065233 0.007 Correlation: (Intr) Weight -0.976 Standardized Within-Group Residuals: Min Q1 Med Q3 Max -1.74326719 -0.41379593 -0.06508451 0.39578734 2.27406649 Number of Observations: 114 Number of Groups: 19 As you can see the p-value for factor Weight is significant. This is the second model, in which I add various factors for searching their correlations: library(nlme) summary(lme(Mode ~ Weight*Height*Shoe_Size*BMI, data = Gravel_ds, random = ~1 | Subject)) Linear mixed-effects model fit by REML Data: Gravel_ds AIC BIClogLik 1975.165 2021.694 -969.5825 Random effects: Formula: ~1 | Subject (Intercept) Residual StdDev:1.127993 2494.826 Fixed effects: Mode ~ Weight * Height * Shoe_Size * BMI Value Std.Error DFt-value p-value (Intercept) 5115955 10546313 95 0.4850941 0.6287 Weight -13651237 6939242 3 -1.9672518 0.1438 Height -18678 53202 3 -0.3510740 0.7487 Shoe_Size 93427213737 3 0.4371115 0.6916 BMI -13011088 7148969 3 -1.8199949 0.1663 Weight:Height 28128 14191 3 1.9820883 0.1418 Weight:Shoe_Size 351453186304 3 1.8864467 0.1557 Height:Shoe_Size -783 1073 3 -0.7298797 0.5183 Weight:BMI 19475 11425 3 1.7045450 0.1868 Height:BMI 226512118364 3 1.9136867 0.1516 Shoe_Size:BMI 329377190294 3 1.7308827 0.1819 Weight:Height:Shoe_Size -706 371 3 -1.9014817 0.1534 Weight:Height:BMI-10963 3 -1.7258742 0.1828 Weight:Shoe_Size:BMI -273 201 3 -1.3596421 0.2671 Height:Shoe_Size:BMI-5858 3200 3 -1.8306771 0.1646 Weight:Height:Shoe_Size:BMI 2 1 3 1.3891782 0.2589 Correlation: (Intr) Weight Height Sho_Sz BMIWght:H Wg:S_S Hg:S_S Wg:BMI Hg:BMI S_S:BM Wg:H:S_S W:H:BM W:S_S: H:S_S: Weight -0.895 Height -0.996 0.869 Shoe_Size -0.930 0.694 0.933 BMI -0.911 0.998 0.887 0.720 Weight:Height0.894 -1.000 -0.867 -0.692 -0.997 Weight:Shoe_Size 0.898 -0.997 -0.873 -0.700 -0.999 0.995 Height:Shoe_Size 0.890 -0.612 -0.904 -0.991 -0.641 0.609 0.619 Weight:BMI 0.911 -0.976 -0.887 -0.715 -0.972 0.980 0.965 0.637 Height:BMI 0.900 -1.000 -0.875 -0.703 -0.999 0.999 0.999 0.622 0.973 Shoe_Size:BMI0.912 -0.992 -0.889 -0.726 -0.997 0.988 0.998 0.649 0.958 0.995 Weight:Height:Shoe_Size -0.901 0.999 0.876 0.704 1.000 -0.997 -1.000 -0.623 -0.971 -1.000 -0.997 Weight:Height:BMI -0.908 0.978 0.886 0.704 0.974 -0.982 -0.968 -0.627 -0.999 -0.975 -0.961 0.973 Weight:Shoe_Size:BMI-0.949 0.941 0.928 0.818 0.940 -0.946 -0.927 -0.751 -0.980 -0.938 -0.924 0.9350.974 Height:Shoe_Size:BMI-0.901 0.995 0.878 0.707 0.998 -0.992 -1.000 -0.627 -0.960 -0.997 -0.999 0.9990.964 0.923 Weight:Height:Shoe_Size:BMI 0.952 -0.948 -0.933 -0.812 -0.947 0.953 0.935 0.747 0.985 0.946 0.932 -0.943 -0.980 -0.999 -0.931 Standardized Within-Group Residuals: Min Q1 Med Q3 Max -2.03523736 -0.47889716 -0.02149143 0.41118126 2.20012158 Number of Observations: 114 Number of Groups: 19 This time the p-value associated to Weight is not significant anymore. Why?