simr
simr copied to clipboard
pitakakariki
powerSim freezing
I have been using powerSim on my data and it is working perfectly when I am looking at the power of categorical variables with two or three levels to them, however continuous variables and a categorical variable with 14 levels cause powerSim to freeze and will result in an error message from R. I was wondering if anyone had a solution for this? I have a large data set with around 14,000 results for each variable.
Thank you.
Sorry this is not an answer. I am shamelessly using the discussion here to ask a question that I asked about a week ago with no feedback yet: https://github.com/pitakakariki/simr/issues/91.
As powerSim is working perfectly for you with categorical variables, could I ask a question that relates to this? I have a linear mixed effects model with two categorical variables (3 and 4 levels) and their interaction in the fixed effects. I am trying to estimate the power for this two-way interaction term, but I always have an error. Do you have an idea how to fix this? Thanks in advance!
powerSim(fit, test=fixed("SconditionScleftO:TargetTypecontrol"),nsim = 30) [ I have deleted the NA datapoints.] Power for predictor 'SconditionScleftO:TargetTypecontrol', (95% confidence interval):============================| 0.00% ( 0.00, 11.57) Test: unknown test Effect size for SconditionScleftO:TargetTypecontrol is 0.13
Based on 30 simulations, (0 warnings, 30 errors) alpha = 0.05, nrow = 5535
Time elapsed: 0 h 0 m 16 s
nb: result might be an observed power calculation Warning message: In observedPowerWarning(sim) : This appears to be an "observed power" calculation
Hello, I have removed my interactions from my model as I found it resulted in a similar issue, even when they weren't the variables I was testing the power for, so I would also be happy to have an answer to your question as it's another issue I've faced!
Hi Thank you so much for your prompt reply! This is reallly helpful. Yes, when I removed the interaction and tested the power for the single effects one by one, powerSim returned a reasonable power but with a warning about "observed power" calculation. Did you have a similar warning as well?
This is my code and the output: fit<-lmer(rttrans~Scondition+TargetType+(1|ID)+(1|probe),data=pilot) powerSim(fit, test=fixed("TargetType"),nsim = 30) Power for predictor 'TargetType', (95% confidence interval):=====================================================| 100.0% (88.43, 100.0)
Test: Likelihood ratio
Based on 30 simulations, (0 warnings, 0 errors) alpha = 0.05, nrow = 5535
Time elapsed: 0 h 0 m 24 s
nb: result might be an observed power calculation Warning message: In observedPowerWarning(sim) : This appears to be an "observed power" calculation
Hi if you make the interaction a single term using "df$AB = interaction(df$A, df$B)", the problem should be solved!
Are you sure it's freezing? Try e.g. nsim=3 to see if it's just taking a long time.
What is the error message you are getting?
Hi everyone,
I guess I'll also use this thread to ask a related question. running powerSim is freezing. Even when i try to run 2 simulations it doenst work. I have a model that looks like this:
Linear mixed model fit by maximum likelihood . t-tests use Satterthwaite's method ['lmerModLmerTest']
Formula: logratio ~ status.numeric * verb.numeric + (verb.numeric + status.numeric ||
subj) + (status.numeric || item)
Data: data3
AIC BIC logLik deviance df.resid
14017.8 14091.9 -6998.9 13997.8 12182
Scaled residuals:
Min 1Q Median 3Q Max
-2.98777 -0.59118 -0.01071 0.77250 2.38581
Random effects:
Groups Name Variance Std.Dev.
item status.numeric 0.0070597 0.08402
item.1 (Intercept) 0.0055717 0.07464
subj status.numeric 0.0022048 0.04696
subj.1 verb.numeric 0.0009728 0.03119
subj.2 (Intercept) 0.0120178 0.10963
Residual 0.1804459 0.42479
Number of obs: 12192, groups: item, 36; subj, 12
Fixed effects:
Estimate Std. Error df t value Pr(>|t|)
(Intercept) 0.067156 0.034229 15.648670 1.962 0.0678 .
status.numeric 0.048668 0.019879 32.816828 2.448 0.0199 *
verb.numeric -0.020590 0.009818 11.248360 -2.097 0.0593 .
status.numeric:verb.numeric -0.064259 0.003916 12155.395970 -16.408 <0.0000000000000002 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Correlation of Fixed Effects:
(Intr) stts.n vrb.nm
status.nmrc -0.001
verb.numerc 0.002 -0.002
stts.nmrc:. -0.002 0.011 -0.021
I wanna run a simulation on the interaction term:
power_1 <- powerSim(m.onset1.sim, fixed("status.numeric:verb.numeric"),nsim = 2)
but the progress bar doesnt move at all and my computer starts purring :) Any ideas on what I'm doing wrong?
I'd start by running through a single simulation to see which step your computer is struggling with:
y <- doSim(m.onset1.sim)
z <- doFit(y, m.onset1.sim)
doTest(z, fixed("status.numeric:verb.numeric"))
If the model fitting in lme4 is working fine, then I suspect it will be the testing step that's the problem. In that case you might want to specify a less computationally intensive test e.g.:
fixed("status.numeric:verb.numeric", "z")
I have been using powerSim on my data and it is working perfectly when I am looking at the power of categorical variables with two or three levels to them, however continuous variables and a categorical variable with 14 levels cause powerSim to freeze and will result in an error message from R. I was wondering if anyone had a solution for this? I have a large data set with around 14,000 results for each variable.
Thank you.
Hello,
I had the same problem and I fixed it by removing the NA values:
df <- df[!is.na(df$var1),]
model <- lmer(dv ~ var1 + var2 + (1|subject) + (1|item), data = df)
powerSim(model, test=fixed("var1"))
