Problems with drawing graph of MSE error for ridge model in R

I built the optimal ridge model:

library(glmnet)

data(Hitters, package = "ISLR")

x <- Hitters[, c("AtBat", "Hits", "HmRun", "Runs", "RBI", "Walks", "Years", "CAtBat", "CHits", "CHmRun", "CRuns", "CRBI", "CWalks", "PutOuts", "Assists", "Errors")]
y <- Hitters$Salary

x <- scale(x)

lambda_seq <- 10^seq(10, -2, length = 100)

ridge_model <- glmnet(x, y, alpha = 0, lambda = lambda_seq)

cv_ridge <- cv.glmnet(x, y, alpha = 0)

lambda_optimal <- cv_ridge$lambda.min

ridge_model_optimal <- glmnet(x, y, alpha = 0, lambda = lambda_optimal)

summary(ridge_model_optimal)

and I would like to draw a bar graph showing the MSE errors. I try to do it with the function:

x_train <- model.matrix(Salary ~ ., data = train)[,-1]
y_train <- train$Salary
x_valid <- model.matrix(Salary ~ ., data = valid)[,-1]
y_valid <- valid$Salary

mse_ridge <- caret::RMSE(predict(ridge_model_optimal, newx = x_valid), y_valid)

but I receive this error:

>Error in predict.glmnet(ridge_model_optimal, newx = x_valid) :
The number of variables in newx must be 16

Do you know how can I fix it?

You haven't shown in the code how you created the train and valid data sets, yet I suspect this is exactly where your problem lies.

First let us load the data and limit ourselves to complete cases:

library(glmnet)

data(Hitters, package = "ISLR")

Hitters <- Hitters[complete.cases(Hitters), ]

Now we can create our x and y data:

x <- Hitters[,c("AtBat", "Hits", "HmRun", "Runs", "RBI", "Walks",
               "Years", "CAtBat", "CHits", "CHmRun", "CRuns", "CRBI",
               "CWalks", "PutOuts", "Assists", "Errors")]

x <- scale(x)

y <- Hitters$Salary

We can create our model like this:

lambda_seq <- 10^seq(10, -2, length = 100)

ridge_model <- glmnet(x, y, alpha = 0, lambda = lambda_seq)

cv_ridge <- cv.glmnet(x, y, alpha = 0)

lambda_optimal <- cv_ridge$lambda.min

ridge_model_optimal <- glmnet(x, y, alpha = 0, lambda = lambda_optimal)

Now let us take samples of Htters to create a training and validation subset at random, with a 2:1 split

set.seed(1)
train_test <- sample(1:2, nrow(x), TRUE, prob = 2:1)

train <- as.data.frame(cbind(Salary = y[train_test == 1], x[train_test == 1,]))
valid <- as.data.frame(cbind(Salary = y[train_test == 2], x[train_test == 2,]))

And we can get the x and y values for train and valid like so:

x_train <- model.matrix(Salary ~ ., data = train)[,-1]
y_train <- train$Salary
x_valid <- model.matrix(Salary ~ ., data = valid)[,-1]
y_valid <- valid$Salary

Now we can get the RMSE however we like, be it via caret or a simple manual calculation:

caret::RMSE(predict(ridge_model_optimal, newx = x_valid), y_valid)
#> [1] 389.107

sqrt(mean((predict(ridge_model_optimal, newx = x_valid) - y_valid)^2))
#> [1] 389.107

You say that you want to plot the RMSE, but it's not clear what you mean by that, since there is only a single value for RMSE. Perhaps you want a histogram of predicted versus actual?

hist(predict(ridge_model_optimal, newx = x_valid) - y_valid,
     main = "Residual histogram", xlab = "Predicted - Actual")

Or maybe show the individual errors?

plot(x_valid[,"AtBat"], y_valid, xlab = "At Bat (normalized)",
     ylab = "Salary", main = "Actual (black) versus predicted (red)")

points(x_valid[,"AtBat"], predict(ridge_model_optimal, newx = x_valid),
       col = "red")

segments(x_valid[,"AtBat"], y_valid, col = "red",
         y1 = predict(ridge_model_optimal, newx = x_valid))