Linear Regression and Logistic Regression using R Studio

Identify the two main data types: qualitative and quantitative, and distinguish their subtypes, such as nominal, ordinal, discrete, and continuous, to guide appropriate analyses.

Celebrate reaching this milestone and stay motivated to complete the course, as you join the top 50% of learners, rate the course, and access Q&A, AI assistant, and a certificate.

Explore descriptive and inferential statistics, including measures of center and dispersion, frequency distributions, bar charts, and histograms, with neural networks covered as a key deep learning technique.

Explore frequency distributions, bar charts, histograms, and relative frequency; compare qualitative and quantitative data and grasp normal distribution, skewness, and uniformity.

Explore descriptive measures of center in data, including mean, median, mode, and mid-range, with examples and distinctions between population mean and sample mean.

Explore measures of dispersion, including range, standard deviation, and variance, and how variance equals the square of the standard deviation. Note that the range is influenced by outliers.

Install R first, then R Studio on Windows, then complete a quick crash course in statistics using the R Studio interface and scripts.

Learn basic R and R studio tasks: write and run commands, comment with #, assign with <-, create vectors with c() or 1:10, manage workspace with ls and rm.

Learn to install, load, and manage packages in R, using lib linear for linear regression and ggplot for plotting. Explore built-in packages and repository installations for reproducible code.

Identify and load built-in data sets in R using the data sets package, inspect iris with str and help, and bring the data into the workspace for analysis.

Explore manual data entry in R: assign values to variables using direct, concatenation, and sequence methods, and input values with the scan function.

Import data from tab-delimited text and comma-delimited csv into R Studio using read.table and read.csv, with headers, revealing product (1862 obs, 4 vars) and customer (793 obs, 9 vars) datasets.

Create bar plots in R to visualize region-based frequency distributions, order and orient bars, customize color and borders, add titles and axis labels, and export images for presentations.

Create a histogram in R Studio using hist to plot age distributions, customize breaks and categories, set frequency, color, labels, and export the chart.

Identify the business context and the factors that impact the variables of interest. Gather primary and secondary research to define variables like cart abandonment and choose a model with data.

Explore data by identifying internal and external data needs, requesting data from stakeholders, and performing quality checks. Use cart abandonment and channel insights to define variables and tidy the data.

Examine a real estate pricing dataset: 506 observations across 19 variables, with a data dictionary and primary key, and learn how to join sources and define variables for regression analysis.

Import the house pricing dataset into R Studio with df <- read.csv and header = true, then view df and str(df) to see 506 observations across 19 variables.

Explore univariate analysis using descriptive statistics to summarize single variables, including mean, median, mode, dispersion and quartiles, and learn how the Extended Data Dictionary reveals outliers and missing values.

Perform univariate analysis in R using the extended data dictionary, plot histograms and bar charts, and identify outliers, skewness, missing values, and useless variables for modeling.

Identify and treat outliers in data using box plots, scatter plots, and histograms. Apply capping and floating with percentile-based or sigma methods to improve prediction accuracy in regression models.

Cap outliers in n_hot_rooms at three times the 99th percentile and floor rainfall at 0.3 times the first quantile using quantile in R, improving mean and median alignment.

Learn how to handle missing values in data for regression by deciding to delete observations or impute with values like mean, median, mode, or zero when sensible, plus segment means.

Impute missing values in R by replacing NA with the mean, using na.rm=TRUE, identify NA positions with is.na, apply the substitution, and verify no NA remains.

Explore seasonality in time-based data and learn how to remove it by computing month factors from yearly and monthly means, then normalize sales for better model fit.

Examine bivariate analysis with scatterplots and correlation matrices to assess two-variable relationships and decide on linearity or transformations, and apply variable transformations to improve model fit and handle multicollinearity.

Transform the crime rate with a log (add one) to achieve a linear relationship with price, and create average distance from dist1–dist4, removing bus terminal after EDD.

Apply univariate and bi variate analysis to keep variables, removing single-value ones like bus terminal. Consider regulatory constraints; use business knowledge and scatterplots to confirm relationships and iterate post-regression.

Create dummy variables to convert non-numeric categorical data, such as airport yes/no and favorite subjects, into 0/1 indicators for regression, using n minus one variables to avoid implying order.

Create dummy variables in R using the dummies package to convert categorical fields like airport and water body into numerical indicators for regression, keeping one fewer dummy than categories.

Explore correlation matrices and coefficients to identify positive, negative, and near-zero relationships, distinguish correlation from causation, and address multicollinearity by selecting the most informative variables.

Use the correlation matrix in R to identify price-related variables, round decimals for readability, and remove parks due to high correlation with air quality to avoid multi collinearity.

Explore how linear regression serves as a simple, foundational tool for supervised learning. Learn the least squares approach and apply it to house price prediction to estimate variable effects.

Apply simple linear regression with one predictor to model y as beta0 plus beta1 x, estimate beta0 cap and beta1 cap via least squares, and interpret residuals and RSS.

Assess the accuracy of sample beta0 and beta1 using residual standard error and confidence intervals to test for a relationship between house price and number of rooms.

Assess model accuracy by comparing predicted and actual values with residual standard error and R squared, examine TSS and RSS, and consider adjusted R squared for model complexity.

Learn to run a simple linear regression in R using lm, with price as dependent variable and room_num as independent variable from df, and interpret beta, p-value, and R-squared.

Extend simple regression to multiple linear regression with 16 predictors, interpreting each beta as the effect of a unit change while others fixed, covering RSS, R-squared, and p-values.

Use the f statistic in multiple regression to test whether the model’s predictors jointly relate to the response, with a 5% p-value threshold to guard against false significance.

Transform categorical variables into dummy variables in linear regression, and interpret coefficients and p-values to assess impact on house price; airport and water body illustrate baseline contrasts.

Learn how to run a multiple linear regression in R using lm with multiple predictors, and interpret coefficients, p-values, r-squared and adjusted r-squared for house price insights.

Split data into training and test sets to evaluate training error and test mean squared error, minimize test error, and guide model selection with validation set, leave-one-out, and k-fold cross-validation.

Explain the bias-variance trade-off in regression, showing how increasing model flexibility lowers bias but raises variance, and how the total test error is minimized at an optimal balance.

Split data into train and test sets using ca tools in R, train a linear model on the training data, and compare mean squared error on train and test data.

Learn three lightweight classification models—logistic regression, k-nearest neighbours, and linear discriminant analysis (LDA)—for predicting binary outcomes, using a preprocessed dataset of 506 property transactions where sold equals 0 or 1.

Import data into a variable named df with read.csv and header = TRUE, adjust backslashes to forward slashes, then view the structure with str(df) to see variables and data types.

Explore two business questions in regression: prediction and inferential, using house data to predict sale within three months and to estimate each variable's impact on the response.

Explore why linear regression cannot be used for classification—issues with multi-level responses, dummy variables, probability interpretation, and sensitivity to outliers—then see how logistic regression overcomes them.

Explore logistic regression for predicting default probability using a sigmoid function, with balance and income as predictors, and learn maximum likelihood estimation to fit the model on credit default data.

Train a simple logistic model in R with one predictor, interpret beta0 and beta1, and assess significance via standard error, z value, and p-value against a threshold.

Using a one-predictor logistic model, we estimate beta0 = 6.1 and beta1 = -0.03, and a p-value of 0.0006 shows price impacts the response.

Extend logistic regression to multiple predictors, estimate all betas by maximum likelihood, and compute probabilities to classify outcomes using a 0.5 boundary.

Build a logistic regression with multiple predictors in Python using sklearn and statsmodels, using all features except the sold variable as X and sold as y; fit and inspect coefficients.

Learn to use a confusion matrix to compare true versus predicted classes and identify type one and type two errors—false positives and false negatives in R Studio.

Analyze classifier performance using confusion matrices, exploring true positives, false positives, true negatives, and false negatives, precision, specificity, and sensitivity. Learn to interpret ROC curves and AUC to compare models.

Compute predicted probabilities from the glm model with predict(type='response'), assign yes/no classes using a 0.5 threshold, and evaluate the results with a confusion matrix.

Split data into training and test sets to evaluate model accuracy on unseen data using a confusion matrix and validation set, leave-one-out, and k-fold cross-validation.

Split data into 80% training and 20% test with a seed, train logistic regression, and evaluate with a confusion matrix on the test set.

Celebrate reaching the final milestone in the linear regression and logistic regression course using R Studio and receive a certificate of completion by email.