Problem running mincLm with covariates

[christianlimberger]

I am trying to run mincLm, but I'm encountering an error that only occurs when I include covariates. This is strange because the same error is happening with other codes that previously worked perfectly.

> df_fdg <- read.csv("final_merge-model.csv")
> df_fdg$APOE4 <- as.factor(df_fdg$APOE4)
> df_fdg$PTGENDER <- as.factor(df_fdg$PTGENDER)
> df_fdg$ABETA <- as.numeric(df_fdg$ABETA)
> df_fdg$PTAU <- as.numeric(df_fdg$PTAU)
> df_fdg$AGE <- as.numeric(df_fdg$AGE)
> df_fdg_CU <- filter(df_fdg, DX == "CU")
> df_fdg_CU$AGE <- scale(df_fdg_CU$AGE)[,1]
> df_fdg_CU$ABETA <- scale(df_fdg_CU$ABETA)[,1]
> df_fdg_CU$PTAU <- scale(df_fdg_CU$PTAU)[,1]
> df_fdg_CU$snp <- factor(df_fdg_CU$snp, levels = c("noncarrier", "carrier"))
> lm1 <- mincLm(FDG_path ~ snp + APOE4 + PTGENDER + AGE + ABETA + PTAU, data = df_fdg_CU, mask = "adni_space.mnc")
Error in formula[[3]][[nTerm]] > 1 :
comparison (>) is not possible for language types

[gdevenyi]

First, you can directly use the scale() function inside the RMINC equation, which is probably more clear for the latest outputs.

I suggest you try each covariate individually, something about the structure of one of your columns isn't as expected.

[christianlimberger]

Thank you Gabriel, I already tried putting scale() inside the function, and also testing each covariate individually, but only got the function running without any covariate...

[gdevenyi]

ok, something about your data is very messed up then. Please provide some kind of view of the df_fdg_CU object.

[christianlimberger]

[gdevenyi]

Ok. Lets look at a few more things.

Given this model is your goal,

> lm1 <- mincLm(FDG_path ~ snp + APOE4 + PTGENDER + AGE + ABETA + PTAU, data = df_fdg_CU, mask = "adni_space.mnc")

Can you successfully run

>  lm1 <- mincLm(FDG_path ~ <SINGLE PREDICTOR>, data = df_fdg_CU, mask = "adni_space.mnc")

For each of these columns successfully?

[christianlimberger]

yes! The problem is when I put any covariate

[gdevenyi]

Okay, lets pare down this to a minimum failing example.

Can you please choose a simple two-predictor model, run it without the mask option, and after the error, run traceback() and provide the full output.

In addition, please provide your sessionInfo()