Skip to content

Setting start values for log binomial model in glmmPQL

4 messages · Ben Bolker, Thierry Onkelinx, Jennifer Yourkavitch

Original
Jennifer Yourkavitch
# 
Hello,
I'm having trouble setting starting values and   I think there is a problem with my syntax for the ?start? command.


First, I?m trying to use the coefficients from a logit model as start values for the log binomial model I want to run:

coefini=coef(glmmPQL(anybf~n_dis +mage_r+mage_sq+married+parity_r+nhblack+hisp+nhasian+mom_lths+mom_sclg+mom_ba+mwork+medicaid+nodad+mom_foreign,
     random=~1|grptr0010,family = binomial))
summary(coefini) 

#coefini model looks good and runs

fit2<-glmmPQL(anybf~n_dis +mage_r+mage_sq+married+parity_r+nhblack+hisp+nhasian+mom_lths+mom_sclg+mom_ba+mwork+medicaid+nodad+mom_foreign,
     random=~1|grptr0010,family = binomial(log),start=c(coefini,1, 0))

summary(fit2)

I get this error: 
Error in glm.fit(x = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,  :
  length of 'start' should equal 16 and correspond to initial coefs for c("(Intercept)", "n_dis", "mage_r", "mage_sq", "married", "parity_r", , "nhblack", "hisp", "nhasian", "mom_lths", "mom_sclg", "mom_ba", , "mwork", "medicaid", "nodad", "mom_foreign")

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
So then I tried typing in the coefficients from the logit model directly:

fit2<-glmmPQL(anybf~n_dis +mage_r+mage_sq+married+parity_r+nhblack+hisp+nhasian+mom_lths+mom_sclg+mom_ba+mwork+medicaid+nodad+mom_foreign,
              random=~1|grptr0010,family = binomial(log),start=c(0.0675, -0.127, 0.008, -0.0002, 0.369, -0.111, -0.0795, 0.363, 0.1387, -0.121, 0.415, 0.861, 0.023, -0.033, -0.307, 0.915))
summary(fit2)

And I get this error:
Error: cannot find valid starting values: please specify some


Can you see the problem?

Thanks!
Jennifer
Ben Bolker
# 
This is a perfectly reasonable question, but it would be easier to
answer with a reproducible example.  See e.g

http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example

On Sun, May 1, 2016 at 5:20 PM, Jennifer Yourkavitch via
R-sig-mixed-models <r-sig-mixed-models at r-project.org> wrote:
Thierry Onkelinx
# 
Dear Jennifer,

Are you sure that the result of c(coefini,1, 0) is a vector of length 16?
The output of c() could be something else than a vector.

Best regards,

ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature and
Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium

To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to say
what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of data.
~ John Tukey

2016-05-01 23:20 GMT+02:00 Jennifer Yourkavitch via R-sig-mixed-models <
r-sig-mixed-models at r-project.org>:

  
  


  


      

    

    

      
      

        


  

        

      Jennifer Yourkavitch
    

    

      
      #
    

  


  

      
Yes, I think you're right. A friend pointed out that I need to specify the coefficients for the fixed effects only.
Thanks,
Jennifer