[R-sig-ME] Can I compare different models - with and without random effects - across different packages?

[David Duffy]

On Thu, 14 Mar 2013, Antonio P. Ramos wrote:

> I see:  perhaps should just keep it there. the problem is that random
> effects makes the computation more complex so that I have less options to
> explore time trends. thanks
>
>
> On Wed, Mar 13, 2013 at 4:45 PM, Philippi, Tom <tom_philippi at nps.gov> wrote:
>
>> The short answer is no, at least for the with & without random effects.
>>
>> On Wed, Mar 13, 2013 at 3:57 PM, Antonio P. Ramos <
>> ramos.grad.student at gmail.com> wrote:
>>
>>> I am trying to see whether I should include a random effects in my models.
>>> I am trying to predict child mortality based on a buch of covariates.

Since your families are small, you can compare results to those from 
a vector glm (VGAM package), as well as the marginal GEE models (gee 
package).

Personally, I test if the within-cluster correlation is zero on a regular 
basis ;).  It's just the distribution of the test that is confusing.

| David Duffy (MBBS PhD)                                         ,-_|\
| email: davidD at qimr.edu.au  ph: INT+61+7+3362-0217 fax: -0101  /     *
| Epidemiology Unit, Queensland Institute of Medical Research   \_,-._/
| 300 Herston Rd, Brisbane, Queensland 4029, Australia  GPG 4D0B994A v