Lära Making Predictions with Models | Statistical Modeling in R

Once you have fitted a statistical model in R, such as a linear regression with lm() or a logistic regression with glm(), you often want to use this model to predict outcomes for new observations. The predict() function in R allows you to generate these predictions by applying your fitted model to a new data frame containing the relevant predictor variables. This is an essential step in using your model for real-world decision making, whether you are forecasting values or classifying new cases.


              1234567891011121314151617
            
# Fit a linear regression model
model_lm <- lm(mpg ~ wt + hp, data = mtcars)

# Fit a logistic regression model (predicting automatic vs. manual transmission)
model_glm <- glm(am ~ mpg + wt, data = mtcars, family = binomial)

# Create a new data frame with predictor values
new_data <- data.frame(wt = c(2.5, 3.2), hp = c(110, 150), mpg = c(21, 18))

# Predict with the linear model
predicted_mpg <- predict(model_lm, newdata = new_data)

# Predict with the logistic model (probabilities)
predicted_prob <- predict(model_glm, newdata = new_data, type = "response")

print(predicted_mpg)
print(predicted_prob)

The predict() function takes your fitted model object as its first argument and a newdata data frame containing the same predictor variables used in model fitting. For example, when using lm(), the predictors in newdata must match those in the formula. For glm() models, you can specify the type argument. Setting type = "response" returns predicted probabilities for classification tasks, while the default returns log-odds. The function returns a vector of predicted values, aligned with the rows in your newdata frame. This allows you to quickly see what your model expects for new cases, given their predictor values.

Note

A common mistake is providing a newdata frame that does not match the variable names or structure used to fit the model.

If column names are misspelled or required predictors are missing, predict() will return an error or produce incorrect results;
Always double-check that your newdata includes all necessary variables with the correct names and types before making predictions.

Var allt tydligt?

Tack för dina kommentarer!

Avsnitt 3. Kapitel 3

Fråga AI

Fråga vad du vill eller prova någon av de föreslagna frågorna för att starta vårt samtal

Suggested prompts:

Can you explain what the predicted values mean in this context?

How do I interpret the probabilities from the logistic regression model?

What should I do if my new data frame has missing predictor values?

Svep för att visa menyn


              1234567891011121314151617
            
# Fit a linear regression model
model_lm <- lm(mpg ~ wt + hp, data = mtcars)

# Fit a logistic regression model (predicting automatic vs. manual transmission)
model_glm <- glm(am ~ mpg + wt, data = mtcars, family = binomial)

# Create a new data frame with predictor values
new_data <- data.frame(wt = c(2.5, 3.2), hp = c(110, 150), mpg = c(21, 18))

# Predict with the linear model
predicted_mpg <- predict(model_lm, newdata = new_data)

# Predict with the logistic model (probabilities)
predicted_prob <- predict(model_glm, newdata = new_data, type = "response")

print(predicted_mpg)
print(predicted_prob)

Note

A common mistake is providing a newdata frame that does not match the variable names or structure used to fit the model.

If column names are misspelled or required predictors are missing, predict() will return an error or produce incorrect results;
Always double-check that your newdata includes all necessary variables with the correct names and types before making predictions.

Var allt tydligt?

Tack för dina kommentarer!

Avsnitt 3. Kapitel 3