Could not figure out the error message

[yiyunzhao]

Could not figure out the error message

[clementinezimnicki]

Hi! I'm having this same issue. Did you ever figure it out?

[Yufanggg]

Is this issue figured out? I have the same one?

[lkumle]

Hi everyone! Sorry for this very delayed response! We added an FAQ that also covers the error message reported here.

https://lkumle.github.io/power_notebooks/intro/FAQ_mixedpower.pdf

Usually, this error occurs when mixedpower cannot use the model's formula to infer column names in the data set used to fit the model (e.g., here, mixedpower will look for a variable called "longdaylanguage". Separating them by either + or * should solve the issue).

[YesimSem]

Hi,

I am having this problem and my variables are named correctly (as they are in the dataframe). What is the solution for this? Thank you

Edit: My model has 3 way interaction, could that be a reason?

[lisakaly]

m4<-lmer(MRSS~ Time * Group+ subset_ssc+ ( 1|subject),data=df, REML = T)

model <- m4 # which model do we want to simulate power for?
data <- df # data used to fit the model
fixed_effects <- c("Time", "subset_ssc", "Group") # all fixed effects specified in m3
simvar <- "subject" # which random effect do we want to vary in the simulation?

SIMULATION PARAMETERS

steps <- c( 40, 50, 60, 80) # which sample sizes do we want to look at?
critical_value <- 2 # which t/z value do we want to use to test for significance?
n_sim <- 500 # how many single simulations should be used to estimate power?

RUN SIMULATION

power_m4<- mixedpower(model=m4, data=df, fixed_effects=c("Time", "subset_ssc", "Group"),
simvar="subject", steps=c(40,50,60,80), critical_value=2, n_sim=500)

This the console response:
[1] "Estimating power for step:"
[1] 40
[1] "Simulations for step 40 are based on 0 successful single runs"
Error in apply(store_simulations, MARGIN = 1, FUN = mean, na.rm = T) :
dim(X) must have a positive length

How can I fix this problem . My variables are named the same in the data set as in the model. And the solution suggested by Ikumie does not really fit to my case....
Thank you for helping

[lkumle]

Hi,

another common source of this error are problems with identifying if a predictor is between- or within-subject. Data sets simulated by mixedpower keep the balance between groups which requires identifying which predictors are between-subject. This fails if e.g., between-subject variables have missing values.

Feel free to reach out via email if this does not apply to your data and I could have a closer look at your data/code.

[paulwarren]

I am also having this problem. I have checked the suggested solutions above and none of them seem to apply to my dataset. Are there any other issues I should be looking for?

UPDATE: I did a fresh installation of all the packages required to install mixedpower, and it now runs okay. So it seems like there was an issue with an old package that was not being reported as such but resulted in this apply error

[AleAnsani]

Hi everyone! I'm getting the exact same error message, and no fresh installation worked.

Here's something potentially new I've discovered:
If I run the lmer version of my glmer model, the simulations run just fine.

That is:
doesnotwork <- glmer(y ~ treatment * time + (1|ID) + (1|stim), family=Gamma(link="log"), data=data)
works <- lmer(y ~ treatment * time + (1|ID) + (1|stim), data=data)

I've also tried with no link function (i.e., identity link function), maybe the log link modifies something inside the lme4 function?! But no success.

@lkumle did this workaround work for other users as well? Any suggestions?
I'd like to run my simulations on my glmer, as it would be more appropriate.

Thank you for your collaboration!

[lkumle]

Hi @AleAnsani ,

In general, "Simulations for step X are based on 0 successful single runs" and "Error in apply(store_simulations, MARGIN = 1, FUN = mean, na.rm = T) : dim(X) must have a positive length" are unfortunately very uninformative error messages. All it allows to conclude is that something went wrong during the simulation process, but it can arise from many different issues.

In your specific case, I know that other users have had issues when using mixedpower with glmer and family = Gamma(), which unfortunately just does not seem to be stable enough at this point. Sorry for not having any concrete suggestions on how to solve this!