Q: re error message "step 2 (trust with pickle3) failed"

[José Waterton]

Hi,

I'm new to aster models and would greatly appreciate some advice with an issue that I'm having.

For some background, I carried out an experiment investigating how herbivores influence phenotypic selection on emergence timing. Fitness was measured as seed production at the end of several months growth. Two seeds of 16 maternal lines of a plant species were planted inside cages which either let in herbivores or excluded them, with 8 cages per herbivory treatment. I had 8 blocks containing two cages of each treatment which also shared the same spatial arrangement of maternal lines. Each seed was weighed before planting (for inclusion as a covariate) and emergence date was recorded. I am treating maternal lines and block as random effects.

I am trying to test whether the effect of emergence time on fitness differs between herbivory treatments by comparing two models, one in which there is an interaction between emergence time and herbivory treatment and a second model without this interaction term. However, for the model with the interaction as shown below:

brdi.aster3 <- reaster(resp ~ varb + fit:(DaysToGerm * HerbTreatment + SeedMassMG),
                       list(Family = ~ 0 + fit : Family, Block = ~ 0 + fit: Block),
                       pred=pred, fam=fam, varb, id, root, data = brdi.redata, famlist = brdi.famlist)

I get the following message:

Error in reaster.default(resp ~ varb + fit:(DaysToGerm * HerbTreatment +  : 
  step 2 (trust with pickle3) failed

The nested model without the interaction doesn't give an error message.

I have absolutely no idea what pickle3 is! I'd greatly appreciate any insight into what this error message means, as the information I've seen elsewhere about pickle3 doesn't make much sense to me as a novice in all this.

Many thanks,

José

[Ruth Shaw]

Greetings, José -

I have conferred with Charlie about this. Here are some suggestions: Make sure you are using the latest version of reaster. Also, doublecheck to make sure that emergence timing is being treated as a continuous predictor.  Finally, you might consider treating blocks as fixed. Estimating variance components for random factors is especially demanding of data.

I hope this helps!

Ruth