Forum | VSN International

GenStat - Help regarding analysis of forestry trial

Keith — Wed, 08 Sep 2010 05:27:51 GMT

Hi,

I was wondering if anybody would be able to help me regarding a forestry trial I am trying to analyse. It was originally laid out as a factorial, but now that I have rotation-end data, there are a couple of concerns I have about the original design.

I have attached a word document where I have tried to explain the design and concerns I have. If any one can help, or point me in the right direction I would be very grateful. Likewise if I haven't explained something clearly enough, then let me know.

Regards
Keith

Keith Little
Institute for Commercial Forestry Research
P O Box 100281
Scottsville
Pietermaritzburg
KwaZulu-Natal
South Africa
3209

Tel: +27 33 386 2314 begin_of_the_skype_highlighting +27 33 386 2314 end_of_the_skype_highlighting begin_of_the_skype_highlighting +27 33 386 2314 end_of_the_skype_highlighting
Fax: +27 33 386 8905
Cell: 082 373 1854

ASReml - RE: Specifying the variance-covariance matrix structure.

nicobardol — Mon, 06 Sep 2010 12:23:15 GMT

Hi Damian,

Thanks for your answer. I try the function str( ), which gives me an unique and equal variance for all my random variables, but when I checked my random effects, it was different from what I was expecting.

I found, the correct function for my problem, if someone is interresting in, I used the function group( ), as followed :

num_column = c(3:696)
liste_column = list (column = num_column)

model.asr = asreml(fixed = yield~1, random=~grp(column), group = liste_column, data = data)

where my vector num_column correspond to the columns numbers of my random variables in my data.frame of data.

Sincerly,

Nicolas B.

ASReml - RE: Specifying the variance-covariance matrix structure.

damianCollins — Wed, 01 Sep 2010 13:53:01 GMT

Dear Nicolas,

If I haven't mis-interpreted your question, I think the str(...) function might be possibly helpful here - something along the lines of:

..., random=~str(~M1+M2+M3+M4+....+Mn, vmodel=~idv(n)),...

where "n" here by the number of markers you have.

You can read the description of str(...) in section 4.3.6 "General variance structures" (~page 54) of the asreml-R manual.

HTH,
Damian

ASReml - RE: Specify a rcov structure (newbie question)

GDM — Mon, 30 Aug 2010 11:21:57 GMT

Indeed. Thanks a lot!

Geert

ASReml - Specifying the variance-covariance matrix structure.

nicobardol — Fri, 27 Aug 2010 14:58:56 GMT

Dear all,

I am working on genetic models. I am using R-ASREML to estimate genetics parameters.

I would like to estimate effects of molecular markers (Mi) considered as random effects.
My assumption is that : All markers (each considered as a single random effect) have the same variance and are all independant. So the variance covariance matrix of the total model (G) is an identity matrix with my variance as diagonal.
I know that by default asreml consider the IDD variance structure but only for each random effect. What should I write in my model to explicit the fact that all my random effect have the same variance.

For know I have :
model.asr = asreml(fixed=yield~1, random=~M1+M2+M3+M4+....+Mi, data=data)
my data file is like as follow :

Individuals yield M1 M2 M3 M4 M5 M6 M7 M8 M9 M10
1 116,5 2 0 1 0 0 1 2 0 2 0
2 119,1 1 1 0 0 0 2 0 2 0 2
3 119,6 0 2 1 1 0 0 1 1 1 1
4 116,5 2 0 0 1 1 0 2 0 2 0
5 118,9 2 0 0 1 1 0 1 1 0 2
....

Thanks a lot for your answers,
Regards,

Nicolas Bardol.

ASReml - RE: Test Example Data Sets

Ali — Thu, 26 Aug 2010 23:29:19 GMT

Many Thanks,
Ali

ASReml - RE: Specify a rcov structure (newbie question)

damianCollins — Thu, 26 Aug 2010 12:24:55 GMT

Dear Geert,

I think what you may be after is
ID:cor(rep)
instead of cor(ID):rep
(to specify the compound symmetry model which is equivalent to fitting random ID effects).

HTH
Damian

GenStat - RE: Multiple correspondence analysis

Alex — Thu, 26 Aug 2010 10:47:34 GMT

Dear Guillermo
In the 13th edition a procedure for Multiple Correspondence Analysis (MCORANALYSIS) was added to GenStat. This was in addition to extending Correspondence Analysis and a biplot procedure for both CA and MCA.
If you don't have the 13th edition I should be able to send you the procedures.
Please let me know if you require them.
Regards
Alex

Alex Glaser
Statistical Software Developer
VSN International Ltd
5 The Waterhouse, Waterhouse Street
Hemel Hempstead HP1 1ES (England)
T: +44-(0)1442-450241
F: +44-(0)870-1215653
www.vsni.co.uk
Registered in England & Wales, company number: 4027977.
Registered Office: Hemel Hempstead (as above).

ASReml - RE: further questions on two traits modeling

Arthur — Thu, 26 Aug 2010 10:30:22 GMT

Dear Wh Lee,

You need to write the model line as

afd yw1 ~ mu Trait Tr.byear Tr.yflockad Tr.byear.yflockad ,
!r !{ Tr.animal at(Tr,1).dam !}

The !{ !} stops ASReml from switching the order of the terms
(it prefers to put the smaller term first).

I have also dropped the !f mv
since multivariate analysis handles missing observations automatically
(i.e. internally without fitting the mv term - unless you really do want
estimates of the missing values. [But mv is required if you use the !ASUV qualifier.

Let me know if this doesn't let your job run.

GenStat - Multiple correspondence analysis

guillermohough — Thu, 26 Aug 2010 10:29:04 GMT

BODY { font-family:Arial, Helvetica, sans-serif;font-size:12px; } Has anyone got a procedure for Multiple Correspondence Analysis?
Thank you very much,
Guillermo
-------------------------------
Dr. Guillermo Hough
DESA - ISETA
H. Yrigoyen 931
(6500) - 9 de Julio - Pcia. de Buenos Aires - Argentina
TEL/FAX: ++54 2317-431309
www.desa.edu.ar

Post generated using Mail2Forum (http://www.mail2forum.com)