[vsni.co.uk] Contact us
Author Message
Post new topic

  GenStat  ~  Genstat: Non-linear models

james brown (JIC)
Posted: Mon Feb 18, 2019 6:37 pm Reply with quote
Guest
Dear Peter,

Thank you for your quick reply. I’ve spent much of the last two weeks revising how to do non-linear models in Genstat and I’ve hit a block with this question of estimating one non-linear parameter separately for different datasets but a common value of another non-linear parameter.

The problem might be that I don’t know how to include factors in an EXPRESSION statement – I don’t even know if factors can be included in a Genstat expression… I’ve been trying to work out how I could use FCLASS and EXPRESSION together to achieve this but I haven’t had any success there either.

Any help you can give me would be much appreciated! If it simply isn’t possible, I guess I’m prepared to bet that reviewers for a microbiology journal won’t realise that the slopes of these graphs should be the same…

With best wishes,

James

From: GENSTAT-Request <GENSTAT@jiscmail.ac.uk> On Behalf Of Peter Lane
Sent: 18 February 2019 15:18
To: GENSTAT@JISCMAIL.AC.UK
Subject: Re: Genstat: Non-linear models

James, I don’t think this can be done either menus, or with commands using the Fitcurve command: you will need to use the Fitnonlinear command. This needs a bit more effort, in particular providing initial values for the nonlinear parameters, but it should be possible unless the model doesn’t fit the data well. I’m on holiday until next week, but let me know if you need more guidance.


Peter Lane


On Mon, 18 Feb 2019 at 17:52, james brown (JIC) <james.brown@jic.ac.uk (james.brown@jic.ac.uk)> wrote:
Quote:

Good afternoon Genstatters,

I’m fitting logistic curves to dose-response data: Z = A+C/(1+EXP(-B(X-M)))
Z = observed growth of parasite, A = lower asymptote of logistic curve (normally growth = 0), C = estimate of growth without pesticide, M = ED50 & B = slope parameter.

I have data for a set of fungal strains. I can do most of what I want to do except that I want to fit curves with a common value of B but different values of M for different strains. The problem is that these are both non-linear parameters and Genstat seems to want to estimate either common values of B and M for all strains or separate values of both B and M for each strain.

In the menus, Stats > Regression > Standard Curves gives me the option of setting Group = Strain. Step-by-step, this:
  • Gives common estimates of A, C, B and M for all strains,
  • Gives separate estimates of A and C with common estimates of B and M (first A estimated separately, then A and C both separately),
  • Then estimates both non-linear parameters separately, which isn’t really what I want.


The commands in the input window are:
TERMS log10Dose*Strain
FITCURVE [PRINT=model,summary,estimates,accumulated; CURVE=logistic; CONSTANT=estimate; FPROB=yes] log10Dose
ADD [PRINT=model,summary,estimates,accumulated; FPROB=yes] Strain
ADD [PRINT=model,summary,estimates,accumulated; FPROB=yes] Strain.log10Dose
ADD [PRINT=model,summary,estimates,accumulated; NONLINEAR=separate; FPROB=yes]

And the parameters estimates from the last two commands are:

Estimates of parameters
-----------------------
Parameter estimate s.e.
B -6.144 0.287
M 0.86170 0.00968
C Strain V295L-1 1.545
A Strain V295L-1 0.09750
C Strain V295L-2 1.657
A Strain V295L-2 -0.04056
C Strain V295L-3 1.599
A Strain V295L-3 -0.008739
C Strain WT 1.634
A Strain WT -0.04959

Then

Estimates of parameters
-----------------------
Parameter estimate s.e.
B Strain V295L-1 -5.205 0.587
M Strain V295L-1 0.9747 0.0361
C Strain V295L-1 1.829 0.135
A Strain V295L-1 -0.197 0.129
B Strain V295L-2 -7.488 0.761
M Strain V295L-2 0.8344 0.0162
C Strain V295L-2 1.5521 0.0545
A Strain V295L-2 0.0557 0.0495
B Strain V295L-3 -8.074 0.872
M Strain V295L-3 0.8454 0.0155
C Strain V295L-3 1.4915 0.0522
A Strain V295L-3 0.0820 0.0475
B Strain WT -4.692 0.476
M Strain WT 0.8160 0.0299
C Strain WT 1.6943 0.0906
A Strain WT -0.0597 0.0832

Can anyone suggest a way of tweaking the commands so I can get separate estimates of A, C and M but a common estimate of B?

Thanks and best wishes,

James Brown

Prof James Brown
Crop Genetics Dept, John Innes Centre
Norwich NR4 7UH, England




To unsubscribe from the GENSTAT list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=GENSTAT&A=1





To unsubscribe from the GENSTAT list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=GENSTAT&A=1

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

Display posts from previous:  

All times are GMT
Page 1 of 1
Post new topic

Jump to:  

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You can attach files in this forum
You can download files in this forum