Error in Plot.window(...) : need Finite 'ylim' values

Question

I have a continuous variable whose n=15000 comments and 451 Als. When rotating the qqnorm For the evaluation of normality, I verified that it does not present normality and therefore I applied a logarithmic transformation. However, when rotating the qqnorm with the log-transformed variable, the graph was not plotted and the following error message appeared:

Error in plot.window(...) : need finite 'ylim' values

Below the script in the order explained above:

> summary(data1$microalb)
Min.  1st Qu.   Median     Mean  3rd Qu.     Max.     NA's 
0.0000   0.3474   0.5042   1.3220   0.8426 308.0000      451 

> qqnorm(data1$microalb) # inspecionando a normalidade
> log.microalb=log(data1$microalb) # aplicando transformação logarítmica
> qqnorm(log.microalb, main="Q-Q Plot - Log microalb", xlab="Quantis Teóricos", ylab= "Quantis Observados")

I deleted the Als from the variable, but the error persisted, so the problem is probably not in the missings. What could be preventing qqnorm generation with log-transformed data?

OBS.: I applied the same transformation in other variables and there was no problem, only in this one.

Answer 1

The minimum of data1$microalb is zero. Therefore, log(min(data1$microalb)) = -Inf. Spin summary(log.microalb) to confirm this.

Delete the infinite information from your dataset. A common way to do this is to add 1 to the vector that should be plotted, as log(1) = 0. Therefore,

qqnorm(log.microalb+1, main="Q-Q Plot - Log microalb", 
  xlab="Quantis Teóricos", ylab= "Quantis Observados")

should solve your problem.

If you just want to remove the infinite observations from your data set, run

qqnorm(log.microalb[!is.infinite(log.microalb)])

although this solution is not the most common in the literature. The most used procedure is precisely the previous one, adding 1 to the data.