R help - Apr 2017 - question about the anova() function for deviance analysis

Dear Maintener,
> I have recently had a bad experience with the anova() function.
> Indeed, I wanted to process a deviance analysis between 2 mixed linear
> models and I was really surprise to see that depending on the ordre in
> which I gave my models, the function did not the same thing: once it
> makes the anova of the first model and in the other ordre, it makes a
> deviance analysis.
> I make lots of formations to researchers who are not always very
> comfortable either with R or with some statistical analysis and they
> were really very disturbed by that.
> Could this be possible to "secure" this anova function in order
that
> it does the same thing whatever the order in which the models are given? 
> Hope I was clear enought and many thanks in advance for your regard
> about my question.
> Best regards, Sophie
--
Sophie Dubois
ANASTATS
sophie.dubois at anastats.fr



---
L'absence de virus dans ce courrier ?lectronique a ?t? v?rifi?e par le
logiciel antivirus Avast.
https://www.avast.com/antivirus

	[[alternative HTML version deleted]]

This list is about R programming; your question seems mostly about
statistics, and is therefore off topic here. I suggest you consult a
local statistical expert who *is* comfortable with such statistical
analyses. In general, partitions of sums of squares in statistical
models can depend on the order of model terms, but you have provided
so little information that I doubt that anyone can guess to what
exactly you refer.

Cheers,
Bert


Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )


On Sun, Apr 16, 2017 at 6:46 AM, Sophie Dubois
<sophie.dubois at anastats.fr> wrote:
> Dear Maintener,
>> I have recently had a bad experience with the anova() function.
>> Indeed, I wanted to process a deviance analysis between 2 mixed linear
>> models and I was really surprise to see that depending on the ordre in
>> which I gave my models, the function did not the same thing: once it
>> makes the anova of the first model and in the other ordre, it makes a
>> deviance analysis.
>> I make lots of formations to researchers who are not always very
>> comfortable either with R or with some statistical analysis and they
>> were really very disturbed by that.
>> Could this be possible to "secure" this anova function in
order that
>> it does the same thing whatever the order in which the models are
given?
>> Hope I was clear enought and many thanks in advance for your regard
>> about my question.
>> Best regards, Sophie
>
> --
> Sophie Dubois
> ANASTATS
> sophie.dubois at anastats.fr
>
>
>
> ---
> L'absence de virus dans ce courrier ?lectronique a ?t? v?rifi?e par le
logiciel antivirus Avast.
> https://www.avast.com/antivirus
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

You are sending your email to a whole mailing list of volunteers, not a specific
"maintainer" (and I am not one). However, your assertions convey
unfamiliarity with statistics rather than deficiencies in R, and this mailing
list is not a stats tutoring list. I did a quick Google search and found  [1],
which might help you get started though.

Also, your description should have included a minimal working example, so no it
was not complete. It is difficult to discuss technical questions online without
having very specific examples of what you did and what you expected (and perhaps
why you expected that).  Read the Posting Guide for more advice on using the R
mailing lists.

Don't forget that there are 10000+ contributed packages for use with R, and
among them you may find alternate algorithms that suit your application better. 
Try the "sos" search package or just use Google to find such options.
If course, you have to have some understanding of the theory you are interested
in to find them, but it isn't a good idea to treat any software as a black
box so that should not really be an impediment if you are doing good quality
research.

[1] http://goanna.cs.rmit.edu.au/~fscholer/anova.php
-- 
Sent from my phone. Please excuse my brevity.

On April 16, 2017 6:46:52 AM PDT, Sophie Dubois <sophie.dubois at
anastats.fr> wrote:
>Dear Maintener,
>> I have recently had a bad experience with the anova() function.
>> Indeed, I wanted to process a deviance analysis between 2 mixed
>linear
>> models and I was really surprise to see that depending on the ordre
>in
>> which I gave my models, the function did not the same thing: once it
>> makes the anova of the first model and in the other ordre, it makes a
>> deviance analysis.
>> I make lots of formations to researchers who are not always very
>> comfortable either with R or with some statistical analysis and they
>> were really very disturbed by that.
>> Could this be possible to "secure" this anova function in
order that
>> it does the same thing whatever the order in which the models are
>given? 
>> Hope I was clear enought and many thanks in advance for your regard
>> about my question.
>> Best regards, Sophie
>
>--
>Sophie Dubois
>ANASTATS
>sophie.dubois at anastats.fr
>
>
>
>---
>L'absence de virus dans ce courrier ?lectronique a ?t? v?rifi?e par le
>logiciel antivirus Avast.
>https://www.avast.com/antivirus
>
>	[[alternative HTML version deleted]]
>
>______________________________________________
>R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>https://stat.ethz.ch/mailman/listinfo/r-help
>PLEASE do read the posting guide
>http://www.R-project.org/posting-guide.html
>and provide commented, minimal, self-contained, reproducible code.

On 16/04/2017 9:46 AM, Sophie Dubois wrote:
> Dear Maintener,
>> I have recently had a bad experience with the anova() function.
>> Indeed, I wanted to process a deviance analysis between 2 mixed linear
>> models and I was really surprise to see that depending on the ordre in
>> which I gave my models, the function did not the same thing: once it
>> makes the anova of the first model and in the other ordre, it makes a
>> deviance analysis.
>> I make lots of formations to researchers who are not always very
>> comfortable either with R or with some statistical analysis and they
>> were really very disturbed by that.
>> Could this be possible to "secure" this anova function in
order that
>> it does the same thing whatever the order in which the models are
given?
>> Hope I was clear enought and many thanks in advance for your regard
>> about my question.
Could you please post a simple example?  I am guessing that the models 
you are comparing are from different modelling functions, so different 
anova methods were called (it's the first argument that determines what 
is called), but I'm not at all sure without seeing an example.

It doesn't need to be your original data, in fact simpler is better, but 
it should be code that anyone can run.

Duncan Murdoch