Page 1 of 1
Error: invalid link function specified: NA
Posted: Sat Oct 06, 2018 2:17 pm
by manolo01
Hi,
this is the first time I am using R2MLwiN.
I want to run a model with random effects only. My database is a sub-sample of the Multiple Indicator Cluster Survey developed by UNICEF. By response variable is a binary variable where 1 indicates age of marriage before 18 y.o. and 0 indicates age of marriage after 18 y.o.
This is the code I wrote
Code: Select all
F1 <- logit(agem.18) ~ 1 + (1 | province) + (1 | w.id)
rand.eff <- runMLwiN(Formula = F1, D = "Binomial", data = moz.20)
and I get the following error message
Code: Select all
Invalid link function specified: NA
Error in runMLwiN(Formula = F1, D = "Binomial", data = moz.20)
I am not sure what is causing this error. The response variable is a binary
Code: Select all
> str(moz.20$agem.18)
Factor w/ 2 levels "No","Yes": 2 1 1 1 2 1 1 2 2 2 ...
Does it have anything to do with the denom? I read that when it is not specified, the denom is automatically set to 1.
Thanks for your help
Manolo
Re: Error: invalid link function specified: NA
Posted: Mon Oct 08, 2018 9:21 am
by ChrisCharlton
This appears to be a bug in our code for parsing the model formula where it does not correctly handle a "
." in the response name, for example:
Code: Select all
> library(R2MLwiN)
> data(bang, package = "R2MLwiN")
> (mymodel1 <- runMLwiN(logit(use) ~ 1 + lc, D = "Binomial", data = bang))
-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-
MLwiN (version: 3.02) multilevel model (Binomial)
Estimation algorithm: IGLS MQL1 Elapsed time : 0.6s
Number of obs: 2867 (from total 2867) The model converged after 4 iterations.
Log likelihood: NA
Deviance statistic: NA
---------------------------------------------------------------------------------------------------
The model formula:
logit(use) ~ 1 + lc
Level 1: l1id
---------------------------------------------------------------------------------------------------
The fixed part estimates:
Coef. Std. Err. z Pr(>|z|) [95% Conf. Interval]
Intercept -1.12288 0.08348 -13.45 3.05e-41 *** -1.28650 -0.95926
lcOne_child 0.93275 0.12156 7.67 1.676e-14 *** 0.69450 1.17100
lcTwo_children 1.09250 0.12509 8.73 2.466e-18 *** 0.84733 1.33768
lcThree_plus 0.87224 0.10302 8.47 2.523e-17 *** 0.67033 1.07416
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
---------------------------------------------------------------------------------------------------
The random part estimates at the l1id level:
Coef. Std. Err.
var_bcons_1 1.00000 0.00000
-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-
> bang$use.test <- bang$use
> (mymodel1 <- runMLwiN(logit(use.test) ~ 1 + lc, D = "Binomial", data = bang))
Invalid link function specified: NA
Error in runMLwiN(logit(use.test) ~ 1 + lc, D = "Binomial", data = bang) :
Until this is fixed the workaround would be to rename
agem.18 to remove the "
." character.
Re: Error: invalid link function specified: NA
Posted: Mon Oct 08, 2018 10:44 am
by manolo01
Thank you, Chris,
I renamed the variable as suggested and now it works and I was able to fit my models.
I posted the same question on StackOverFlow, can I share the link to this page as credit for your reply and close the question there?
Cheers
Manolo
Re: Error: invalid link function specified: NA
Posted: Mon Oct 08, 2018 1:29 pm
by ChrisCharlton
I am glad to hear that this is now working for you. Yes, you are welcome to link to this page if you wish. I have now also added a reference to this bug to the R2MLwiN known bugs page (
http://www.bristol.ac.uk/cmm/software/r ... nbugs.html).