The fix, courtesy of Kamil Barton, the author of MuMIn:

it is because 'dredge' does not accept an argument called 'na.action',

so passes it as '...' to AICc that does not know what to with it either.

Either use na.action in your glm call or set it as a global option, i.e:

> glm(..., na.action = na.fail)

or

> options(na.action = na.fail)

I've used MuMIn a lot but haven't encountered this problem. First the obvious (and unlikely) but is one of the first values of your dataset an "NA"?

str(dataframenamehere) 

Second, I think this error is related to your choice of "na.fail". Will the model run if you omit rows of data with missing values:

lipid.model2

Hi Jason

Thanks for the advice. The file I'm working with is huge, therefore the str function provides a truncated output. I've checked the source file for all variables - there are no missing data here.

I've also tried running with na.omit, I get a similar error message.

Catch 22 is that MuMIn won't run if I don't specify an na.function.

I said some choice words at the screen yesterday!

Hi Joe,

Sometimes has happend to me that the functions related to NA (i.e. na.omit, na.fail, na.exclude..) have modified the type of object to be analyzed causing errors. If this has happened to you maybe that is the reason why MuMIn is not working properly. You said you already have checked there are no NAs in your data (You could use is.na() function to check again your data, you will have an “True of False” answer). If the answer is false you were right about NAs and I do know what may be causing the error… (sorry!). If the answer is “True” you could try to remove “NA” by using na.omit , for instance, but make sure the data still preserve the type of object you need (e.g. data frame, matrix …).

Hope it helps

Gema

I guess at this point I wold systematically omit variables from the full model and see if it runs at all and in this process ID the variable that is likely bombing this model.

Thanks Gema - didn't know about the is.na() function so I've learned something there!

Jason - I think that's a good idea. I'll try and generate a simpler source file with only the details I'm interested in and then perform as you suggest. Coffee will likely be my friend!

Thanks both for the suggestions - much appreciated :)

Hi Joseph

Did you manage to figure out what was going on? I'm getting exactly the same error message even though I have no NAs, the glm runs without any problems, and the script previously ran without this error. I can't find any solutions online. I'd really appreciate any tips, this is driving me crazy!

Thanks

The fix, courtesy of Kamil Barton, the author of MuMIn:

it is because 'dredge' does not accept an argument called 'na.action',

so passes it as '...' to AICc that does not know what to with it either.

Either use na.action in your glm call or set it as a global option, i.e:

> glm(..., na.action = na.fail)

or

> options(na.action = na.fail)

Thanks heaps Sally! Will give this a go today - you have no idea how close the computer came to flying through the window!

How's sunny Copenhagen?

Thanks Christian.

I'm still relatively new to R, but am amazed at the number of different ways it's possible to do the same thing!

Dear all,

It's my first post on researchgate. I have the same error message, even though I have any NAs. So i used the command na.action=na.fail in the original lm() model and then used the "dredge" function. It works well now, no error message, nevertheless I don't understand why I must specify this option if I haven't any NA in my data and what is its impact on the model.avg results?

Thanks a lot for your help

I wrote a function to set the parameters correctly and to dredge models in parallel, drastically reducing the time it takes to dredge a model. Here's the script, you can run it by entering the following code (with the correct parameters; notice the quotes around  the data and model object names).

source("http://goo.gl/S5vMNv"); par_dredge (data="data_object_name", model="model_object_name")

Dear Jeroen,

Thanks a lot for your help and your R code.

Kindest regards

You're welcome. Let me know if you run into any problems with the code.

Hi Jeroen,

I'm getting a 404 when I try and look at the source of the code, any chance you changed the location?

Cheers!

Hi All!

Any other tricks? I have tried everything but no success.

mod options(na.action = "na.omit")

> dd

Hi Ezequiel Fabiano ,

I've run into this problem as well, but found that if you duplicate the na.actions, dredge() with throw errors. Try having na.action='na.fail' just in your model and take out the global option.

Cheers,

Theadora