An error occurred when only two or three variables were used as input in the df format

[TBlab-why]

I am using the df format occurrence point and background point as input, and there are only two environment variables. Get the following error：
*** Running initial checks for Castanopsis purpurella_thin1 ... ***
载入需要的命名空间：rJava

• Variable values were input along with coordinates and not as raster data, so no raster predictions can be generated and AICc is calculated with background data for Maxent models.
• Caused by error in ENMeval::ENMevaluate():
  ! * If inputting variable values without rasters, please make sure these values are included in the occs and bg tables proceeding the coordinates.

When the other Settings are unchanged and the environment variables are increased to 4 or more, it will run normally.
ENMeval v2.0.4，maxent.jar v3.4.4
May I ask what causes this？Thank you.

[TBlab-why]

ENMeval v2.0.4，maxent.jar v3.4.4

[jamiemkass]

Can you please show me the code you used to run ENMevaluate()?

[TBlab-why]

`#In this example, occdata has only two variables. Return error.
occdata <- read.csv("D:/Desktop/example/occdata.csv")
bgdata <- read.csv("D:/Desktop/example/bgdata.csv")
e <- ENMeval::ENMevaluate(
occs = occdata,
bg = bgdata,
tune.args = list(fc = "LQ", rm = 1),
partitions = "jackknife",
#其他额外设置，有默认值
taxon.name = "Castanopsis purpurella_thin1",
algorithm = "maxent.jar")

#In this example, occdata has only four variables. Successful operation.
occdata <- read.csv("D:/Desktop/example/occdata1.csv")
bgdata <- read.csv("D:/Desktop/example/bgdata1.csv")
e <- ENMeval::ENMevaluate(
occs = occdata,
bg = bgdata,
tune.args = list(fc = "LQ", rm = 1),
partitions = "jackknife",
#其他额外设置，有默认值
taxon.name = "Castanopsis purpurella_thin1",
algorithm = "maxent.jar")`

[jamiemkass]

You might be leaving out the coordinates. The SWD data frames should have x, y, and then the predictor variable values. I made this more clear in the 2.0.5 documentation. Please confirm whether this is the issue or not.

[TBlab-why]

I am so sorry that I have made such a stupid mistake. Thank you very much for your patient answer.
I have another question for you. I want to choose an appropriate set of environment variables. Keeping fc and rm unchanged and using different environment variables for model tuning, can their results be compared with each other?

[jamiemkass]

Sorry for not answering this before. This is difficult because AUC and other model discrimination metrics may look better just because more variables are included in the model. See my paper about this issue -- I used AIC to compare models with different numbers of predictor variables because it has a built-in penalty term for more variables.