Split a dataframe and apply a GAM to each subset

Question

I have a dataframe that I would like to split into subsets and apply a GAM to each one. Ultimately I'd like to output a set of predictions to one large dataframe.

So far the approach I have tried to take:

• Create function to build the model
• Subset data into list of dataframes
• Use lapply to turn list of dataframes into list of models

Is this the right approach or is there a better way?

# Function to create generalised additive model then create predictions
gam.function <- function(x)  {

  gam.x <- gam(x$switch ~ s(x$input_1) + s(x$input_2),   data=x) #%>%

  #predict(gam.x, newdata = data.frame(input_1 = 3500, input_2 = 13500))

}

# Create a list of data frames for position
position.split <- split(switch,switch$position)

# Return a list of models
model.list <- lapply(position.split, FUN=gam.function) 

An error is returned:

> model.list <- lapply(position.split, FUN=gam.function)
Error in gam(x$switch ~ s(x$input_1) + s(x$input_2),  : 
  Model has more coefficients than data

I know this error seems very clear but I'm struggling to interpret as I can successfully run one subset at a time without lapply.

Answer 1

Your function could be:

gam.function <- function(x)  {
  if (nrow(x) < 3) return(" nrow() to small ")
  gam.x <- gam(switch ~ s(input_1) + s(input_2), data=x) #%>%  
  #predict(gam.x, newdata = data.frame(input_1 = 3500, gas_kwh = input_2))  
}