Lesson 22: Bootstrap confidence intervals for linear regression

April 11, 2018

Review:

• Bootstrap confidence intervals
• Exam 2 Redo on Monday
  • Remember, it’s optional if you’re content with your current score/grade
  • Covers all Python and DataCamp material
  • Format similar to original Exam 2 but perhaps less technical and more conceptual
  • Multiple choice plus some short answer
    • Code snippets you should know how to write
      • For loops (for i in range(100):)
      • If statements (+ if-else and if-elif-else)
      • Random number generators
        • np.random.randint(min, max)
        • np.random.normal(mean, std, n)
        • np.random.choice(1d_variable, n)
      • Summary statistics (e.g., np.mean(), np.median(), np.std())
      • Regression with np.polyfit(x, y, 1)
      • Read csv files with pandas (e.g., pd_variable = pd.read_csv(‘filename.csv’))
      • List operations
        • initialize e.g., list_variable = []
        • append to list e.g., list_variable.append(new_variable))
        • slicing e.g., list_subset = list_variable[2:4]
    • Conceptual material you should understand
      • Monte Carlo simulation (repeated random sampling to generate probability distribution)
      • Bootstrapping (simulation by repeated sampling of original data set)
      • Confidence Intervals
      • Graphics for Exploratory Data Analysis (EDA)
        • histograms, stripplots, boxplots, swarmplots (univariate)
        • scatterplots, trendlines (bivariate)

Presentation:

• Pairs bootstrap for linear regression
  • repeated sampling of 2 (or more) values
  • calculation of model output (y-hat)
  • use np.percentile() to calculate confidence intervals
  • plot multiple trendlines to visualize confidence intervals

Activity:

• Use real estate data to generate selling price estimates with confidence intervals
  • Use Total SqFt and Selling Price
  • Produce numeric estimates
  • Produce graphical representation

Assignment:

• Study for Exam 2 Redo