Convergence error code = 1 message = false convergence (8)-CodePudding

The convergence problem is something I never came to understand entirely, but I as far as I know, there are some keys in the code to get the model to convergence I am not attaching all the database because long-format exceeds the limits here.

I would actually want to ask 2 questions:

How can I skip it in a loop? Meaning: some piece of code to skip this variable in the loop

#sthg like
if(error code = 1) next

I am really intriged, why am I getting results when I run the model individually for the variable is causing trouble?

genes_wide_s18 <- genes_s18 %>% pivot_wider(names_from = gen, values_from = dCt_s18)

genes_wide_s18$time [genes_wide_s18$time == "1"] <- 1 
genes_wide_s18$time [genes_wide_s18$time == "3"] <- 3 

genes_wide_s18$time<-as.numeric(genes_wide_s18$time)


# names of variables
genes.names <- colnames(genes_wide_s18)[25:79]
gene.list$s18 <- NULL
no.genes <- length(genes.names)

# create a named list to hold the fitted models
gene.list <- as.list(1:no.genes)
names(gene.list) <- genes.names

# loop over gene names

for(i in genes.names){ 

  

# print status
  print(paste("Running entity:", i, "which is", which(genes.names==i), "out of", no.genes))
  
# trying to ignore variable because of few observations
 if(i == "s18")next
     print(i)

  
# create temporary data matrix and model formula
  #1st: adjusting per several variables
  tmp <- genes_wide_s18[, c(i,"edad0","time", "grup_int", "id", "sexo", "peso1")]
  fml <- as.formula( paste( i, "~", paste(c("sexo:peso1", "edad0","time", "grup_int", "time:grup_int"), collapse=" ")))

   
  #2nd: adjusting just per group

tmp <- genes_wide_s18[, c(i,"grup_int","time", "id")]
fml <- as.formula( paste( i, "~", paste(c("time:grup_int"), collapse=" ")))
  
gene.list[[i]] <- lme(fml, random= ~ time|id, method="REML", data=tmp, na.action = na.omit)
  
  # assign fit to list by name
 
  
  gene.list[[i]] <- lme(fml, random= ~ time|id, control = lmeControl(opt = "optim"), method="REML", data=tmp, na.action = na.omit)

 
}

.
.
.
.
[1] "Running entity: rxra which is 6 out of 55"
[1] "rxra"
Error in lme.formula(fml, random = ~time | id, method = "REML", data = tmp,  : 
  nlminb problem, convergence error code = 1
  message = false convergence (8)

However when I run this individually, substituting i = rxra in fml :

lme(as.formula( paste( "rxra", "~", paste(c("time:grup_int"), collapse=" "))), random= ~ time|id, control = lmeControl(opt = "optim"), method="REML", data=tmp, na.action = na.omit)

Linear mixed-effects model fit by REML
  Data: tmp 
  Log-restricted-likelihood: -392.218
  Fixed: as.formula(paste("rxra", "~", paste(c("time:grup_int"), collapse = " "))) 
   (Intercept) time:grup_int1 time:grup_int2 time:grup_int3 
     3.6723896     -0.2699351     -0.2630362     -0.2104078 

Random effects:
 Formula: ~time | id
 Structure: General positive-definite, Log-Cholesky parametrization
            StdDev    Corr  
(Intercept) 1.6296951 (Intr)
time        0.7984285 -0.982
Residual    0.5870576       

Number of Observations: 279
Number of Groups: 140

str(tmp)
tibble [15,968 × 4] (S3: tbl_df/tbl/data.frame)
 $ rxra    : num [1:15968] NA NA NA NA NA ...
 $ grup_int: Factor w/ 3 levels "1","2","3": 3 3 3 3 3 3 3 3 3 3 ...
  ..- attr(*, "label")= chr "GENERAL: Grupo de intervención"
 $ time    : num [1:15968] 1 1 1 1 1 1 1 1 1 1 ...

CodePudding user response：

you seem to run lme twice for a given i, immediately overwriting the first result with the second. it looks like when you run it manually you mimic the second call where you specifiy control= ; so this is likely the difference between getting the failure message and not.

  
gene.list[[i]] <- lme(fml, random= ~ time|id, method="REML", data=tmp, na.action = na.omit)
    # assign fit to list by name  
  gene.list[[i]] <- lme(fml, random= ~ time|id, control = lmeControl(opt = "optim"), method="REML", data=tmp, na.action = na.omit)