From: Bruce Southey <bsouthey_at_GMAIL.COM>

Date: Mon, 19 Oct 2009 08:46:41 -0500

Date: Mon, 19 Oct 2009 08:46:41 -0500

Hi,

I can not follow your limited data and no explanation.

If you have survival data then you can an individual die but then

apparently still be alive.

ID A440 appears to die on day=3 yet appears to be alive on day=52.

Either this is an error or you are not really analyzing survival data.

As Arthur indicated, you must first get the data and then the model correct.

Apart from this, it appears that you are using an animal-time period

discrete model (such as I used in my Journal of Animal Science paper).

However, the coding appears to say that '0' is the event indicator

rather than '1'. So please recode this because, to my knowledge, ASREML

can only model the 1 as the event. Also, please use a binary analysis

with the contemporary log-log link function so that you are modeling the

hazard rather than odds.

If you are using a discrete survival model:

1) Average daily gain is a term varying covariate so you can not fit it

like that.

2) Please explain why are you doing a random regression as it may

understand why you are trying to achieve?

Bruce

On 10/16/2009 05:35 AM, Arthur Gilmour wrote:

*> Dear Jose
*

*>
*

*> I suspect there are several forms of survival analysis.
*

*> I understand in your case that each individual is represented by 52
*

*> lines of data and survival switches from 1 to 0 somewheerre through the
*

*> period of 52 days.
*

*>
*

*> Do you have a reasonable number of families (= dams I gather)
*

*>
*

*> Assuming you have a good number of individuals,
*

*> I would probably start by looking at the overall profile of survival
*

*> i.e. ignore genetics to start with. Is death linear with time?
*

*> Some kind of sigmoid curve is more likely is it not.
*

*>
*

*> Then I would add ID to the model since this represents just a mean
*

*> shift. Then add dam to see if there is a genetic component to the
*

*> individual vasriation.
*

*>
*

*> Only if you have a lot of individuals and families would I attempt the
*

*> change of slope at the genetic level.
*

*>
*

*> See below
*

*>
*

*> On Fri, 2009-10-16 at 01:19 +0100, jozzeman wrote:
*

*>
*

*>> Hi,
*

*>>
*

*>> I'm trying to analyse survival data using a random regression model,
*

*>> but I have some problems interpreting the results. The data structure
*

*>> in the data file is as follows:
*

*>>
*

*>> ID father mother GDP1 Tank GDP2 day SR
*

*>> A440 1 3 0.8 1 0.2 1 1
*

*>> A440 1 3 0.8 1 0.2 2 1
*

*>> A440 1 3 0.8 1 0.2 3 0
*

*>> : : : : : : : :
*

*>> : : : : : : : :
*

*>> A440 1 3 0.8 2 0.2 52 0
*

*>> B550 1 3 0.5 2 0.4 1 1
*

*>> B550 1 3 0.5 2 0.4 2 1
*

*>>
*

*>>
*

*>>
*

*>>
*

*>> Reading the data is as follows:
*

*>>
*

*>> Survival analysis
*

*>> ID !P
*

*>> father
*

*>> mother
*

*>> GDP1 !m-9.9 # average daily gain before challenge to disease
*

*>> Tank # 2 levels
*

*>> GDP2 !m-9.9 # average daily gain after challenge to disease
*

*>> day # days of challenge tests (from 1 to 52)
*

*>> SR # survival score at a given day (1 or 0 depending if individual
*

*>> is alive or dead in a specific day)
*

*>> sr2.txt !ALPHA !REPEAT
*

*>> sr2.txt !NODISPLAY !MVINCLUDE !SKIP 1 !maxiter=500
*

*>>
*

*>> I started fitting the follow model:
*

*>>
*

*>> SR ~mu pol(day,1)*Tank !r pol(day,1).mother !f mv
*

*>>
*

*> For this to work, 'mother' should be a 'FACTOR'.
*

*> but I think the model needs ide(ID) first.
*

*>
*

*>
*

*>> 0 1 1
*

*>>
*

*> This shuld be
*

*> 0 0 1
*

*>
*

*>> pol(day,1).mother 2
*

*>> pol(day,1) 0 US .1 .01 1 !GP
*

*>> mother
*

*>>
*

*>> I am interested in estimating the genetic variance of the family (full
*

*>> sibs) this is why I am including only the random effect associated
*

*>> with mother effect.
*

*>>
*

*>> Am I in the right way?
*

*>>
*

*> See above
*

*>
*

*>> See above
*

*>>
*

*>> Can I use !BINOMIAL and !LOGIT qualifiers in the model in order to
*

*>> take account the binary nature of the trait?
*

*>>
*

*>>
*

*> YES but I would concentrate on a plausible model on the 0,1 scale first.
*

*>
*

*>
*

*>> How can I estimate heritabilities for SR?
*

*>>
*

*>>
*

*> Lets get the model sorted, and sensible from a genetic perspective
*

*> first. Then it should be clearer.
*

*>
*

*>
*

*>> How can I fit a multivariate analysis for estimating the genetic
*

*>> correlation between SR and GDP2?
*

*>>
*

*>> I would appreciate very much any suggestions or comments about the
*

*>> coding for the analysis.
*

*>>
*

*>> Thanks in advance!
*

*>>
*

*>> JosÃ©
*

*>>
*

*>>
*

*>>
*

*>>
*

*>> -------------------- m2f --------------------
*

*>>
*

*>> Sent using Mail2Forum (http://www.mail2forum.com).
*

*>>
*

*>> Read this topic online here:
*

*>> http://www.vsni.co.uk/forum/viewtopic.php?p=945#945
*

*>>
*

*>> -------------------- m2f --------------------
*

*>>
*

*>>
*

*>>
*

*>>
*

*>> This message is intended for the addressee named and may contain confidential information. If you are not the intended recipient, please delete it and notify the sender. Views expressed in this message are those of the individual sender, and are not necessarily the views of their organisation.
*

*>>
*

Received on Tue Oct 19 2009 - 08:46:41 EST

This webpage is part of the ASReml-l discussion list archives 2004-2010. More information on ASReml can be found at the VSN website. This discussion list is now deprecated - please use the VSN forum for discussion on ASReml. (These online archives were generated using the hypermail package.)