That is a generous offer. Our guess was that there were not en=

Alain,=0A =0AThat is a generous offer. Our guess was that there were not en=
ough observations to model the varience for each of the cells so we tried m=
odeling with different combinations of the cells but that still didn't help=
. I ran the analysis in 2000-2001, two computers ago, so it may take a whil=
e to find the files and my notes.=0A =0AJohn=0A=0A=0A----- Original Message=
 ----=0AFrom: Highland Statistics Ltd. <[EMAIL PROTECTED]>=0ATo: ECOLOG=
[EMAIL PROTECTED]: Wednesday, March 14, 2007 3:33:56 PM=0ASubject:=
 Re: [ECOLOG-L] Dealing with non-normal, ordinal data for 2-way ANOVA with =
interactions=0A=0A=0A>You seem a trifle sensitive about models and modeling=
 - statistics =0A>are just tools.=0A=0AFully agree......but these tools tha=
t have to be applied correctly.=0A=0A>  It makes no sense to say that somet=
hing is wrong with the data.=0A=0AI see data sets all the time that are wro=
ng; mistakes, typos, wrong =0Acoding of categorical variables, blanks, impr=
oper variable names, =0Aetc, etc. These can make a GLM routine crash.=0A=0A=
>Either the program works for its intened purpose or it doesn't.=0A=0AHere =
is another reason for non-convergence: models with interactions =0Aapplied =
on highly unbalanced data; not enough observations in a =0Aparticular two- =
or three way interaction combination.=0A=0A>If one of the statisticians who=
 helped debug the program for SAS and =0A>another professional statistician=
/programmer cannot get the program =0A>to work with a data set I'd say that=
 the functionality of of the =0A>algorithm depends on the data set - it is =
a tool that sometimes =0A>can't handle the data.=0A=0AImproper application =
of a particular method, over-complicated models =0Awith too many parameters=
,  or "funny" things with the data comes to =0Amy mind first. I have even s=
een people applying a Poisson GLM on =0Anegative data, and wondering why it=
 crashed.=0A=0APerhaps you are willing to upload your data to a website, so=
 that we =0Acan all have a go at it? That would solve this discussion for o=
nce =0Aand for all. Else we will never know why, and how you managed to =0A=
crash the GLM routine in SAS.=0A=0AAlain=0A=0A=0A=0A=0A=0A=0A=0A=0ADr. Alai=
n F. Zuur=0AFirst author of:=0A=0AAnalysing Ecological Data (2007).  Zuur, =
AF, Ieno, EN and Smith, GM. =0ASpringer. 680 p.=0AURL: www.springer.com/0-3=
87-45967-7=0A=0AAnalysing Ecological data using GLMM and GAMM in R. (2008).=
 Zuur, AF, =0AIeno, EN, Walker, N and Smith, GM=0ASpringer.=0A=0AOther book=
s: http://www.brodgar.com/books.htm=0A=0AStatistical consultancy, courses, =
data analysis and software=0AHighland Statistics Ltd.=0A6 Laverock road=0AU=
K - AB41 6FN Newburgh=0ATel: 0044 1358 788177=0AEmail: [EMAIL PROTECTED]
m=0AURL: www.highstat.com=0AURL: www.brodgar.com

Reply via email to