D600 Task 2: Logistic Regression
Analysis
A. GitLab Repository
A subgroup and project were successfully created in GitLab for Task 2. The project was
correctly cloned into the IDE, and commits were made after the completion of each rubric-
aligned section. Each commit included a descriptive message and timestamp to ensure
proper version tracking and traceability. The GitLab repository URL was submitted in the
“Comments to Evaluator” section as instructed, and a full branch history was exported and
included in the final submission.
Section B: Research Question and Goal
B1. Research Question
The research question for this logistic regression analysis is: To what extent do housing
features such as SquareFootage, Garage, RenovationQuality, Price, SchoolRating, and
Fireplace predict whether a home is classified as luxury (IsLuxury)? This question reflects a
realistic business need, such as understanding which factors influence the classification of
luxury homes to inform pricing or investment decisions.
B2. Data Analysis Goal
The goal is to determine the strength and direction of influence that selected housing
attributes have on the likelihood of a home being classified as luxury. By applying logistic
regression, we aim to build a predictive model using physical and economic features,
allowing stakeholders to make data-informed decisions on property evaluation.
,Section C: Data Preparation
C1. Variable Identification
In this analysis, the dependent variable selected is Price, which represents the market value
of a home. This variable is continuous and well-suited for linear regression analysis. The
purpose of the model is to identify which property characteristics significantly influence
housing prices, a scenario relevant to real estate firms, appraisers, and property investors.
To predict Price, six independent variables were selected based on theoretical justification
and practical relevance to home valuation. These include both quantitative and categorical
variables:
SquareFootage (quantitative): Larger homes are expected to command higher
prices, making this a core explanatory variable.
Garage (categorical): Homes with a garage often offer added utility and security,
which may positively influence price.
RenovationQuality (quantitative): The quality of renovation reflects investment in
the home’s condition and appeal, which likely increases its market value.
BackyardSpace (quantitative): More outdoor space typically adds value, especially
in suburban and family-oriented markets.
Fireplace (categorical): Fireplaces are often viewed as a luxury or desirable feature,
potentially adding to a home’s price.
SchoolRating (quantitative): Higher-rated schools are associated with desirable
neighborhoods, often correlating with higher home prices.
These variables were selected to ensure a comprehensive model that includes both
structural and locational features commonly used in real-world housing valuation models.
C2. Descriptive Statistics
To better understand the variables in the dataset, descriptive statistics were computed for
all selected features. For continuous variables (SquareFootage, RenovationQuality,
BackyardSpace, SchoolRating, and Price), the analysis includes the mean, standard
deviation, and the five-number summary: minimum, first quartile (Q1), median, third
quartile (Q3), and maximum.
"In addition to summary statistics, frequency distributions were calculated for the
categorical variables. For example, 'Garage' appeared in approximately 36% of the homes,
while 64% lacked one. Similarly, about 40% of homes had a 'Fireplace'. The target variable,
'IsLuxury', showed a roughly 50/50 distribution, with 3,500 non-luxury and 3,500 luxury
homes out of 7,000 total. These distributions help confirm variable balance and guide
modeling decisions."
, (Screenshots of the descriptive statistics outputs for each variable are included in the
appendix section of the report).
C3. Visualizations
Univariate visualizations were constructed to explore the distribution of each variable
individually. For continuous variables like Price and SquareFootage, histograms were used.
The histogram of Price showed a right-skewed distribution, with a high frequency of homes
priced between $200,000 and $300,000. The histogram for SquareFootage revealed a
normal-like distribution with a few larger homes extending the upper range.
For categorical variables such as Garage and Fireplace, bar charts were used. These
visualizations illustrated the relative proportions of homes with and without each feature.
For instance, the bar chart for Garage showed a higher count of homes without garages,
supporting the descriptive statistics.
Bivariate visualizations were created to assess relationships between each independent
variable and the dependent variable (Price). For continuous predictor variables (e.g.,
SquareFootage, BackyardSpace, RenovationQuality, and SchoolRating), scatterplots were
used to visualize linear relationships. These scatterplots generally showed positive
associations—for instance, as SquareFootage increased, so did Price.
For categorical variables such as Garage and Fireplace, side-by-side boxplots were created
to compare the distribution of Price between the two categories (e.g., Garage = Yes vs. No).
The boxplot for Garage indicated that homes with garages had noticeably higher median
prices than those without, supporting its inclusion as a predictor.
Each visualization is explained with plain-English interpretations in the appendix, and
screenshots of these plots are included to support the visual analysis.
Section D: Data Analysis
D1. Data Splitting
The dataset was split into training (75%) and test (25%) sets using the scikit-learn
`train_test_split` function. Categorical variables were one-hot encoded where applicable,
and quantitative variables were used without scaling, following D600 guidelines.
D2. Model Optimization
The logistic regression model was built using the `statsmodels` package. Backward stepwise
elimination was used to remove variables with p-values above 0.05, one at a time. This
process ensures that only statistically significant predictors remain in the optimized model.
Analysis
A. GitLab Repository
A subgroup and project were successfully created in GitLab for Task 2. The project was
correctly cloned into the IDE, and commits were made after the completion of each rubric-
aligned section. Each commit included a descriptive message and timestamp to ensure
proper version tracking and traceability. The GitLab repository URL was submitted in the
“Comments to Evaluator” section as instructed, and a full branch history was exported and
included in the final submission.
Section B: Research Question and Goal
B1. Research Question
The research question for this logistic regression analysis is: To what extent do housing
features such as SquareFootage, Garage, RenovationQuality, Price, SchoolRating, and
Fireplace predict whether a home is classified as luxury (IsLuxury)? This question reflects a
realistic business need, such as understanding which factors influence the classification of
luxury homes to inform pricing or investment decisions.
B2. Data Analysis Goal
The goal is to determine the strength and direction of influence that selected housing
attributes have on the likelihood of a home being classified as luxury. By applying logistic
regression, we aim to build a predictive model using physical and economic features,
allowing stakeholders to make data-informed decisions on property evaluation.
,Section C: Data Preparation
C1. Variable Identification
In this analysis, the dependent variable selected is Price, which represents the market value
of a home. This variable is continuous and well-suited for linear regression analysis. The
purpose of the model is to identify which property characteristics significantly influence
housing prices, a scenario relevant to real estate firms, appraisers, and property investors.
To predict Price, six independent variables were selected based on theoretical justification
and practical relevance to home valuation. These include both quantitative and categorical
variables:
SquareFootage (quantitative): Larger homes are expected to command higher
prices, making this a core explanatory variable.
Garage (categorical): Homes with a garage often offer added utility and security,
which may positively influence price.
RenovationQuality (quantitative): The quality of renovation reflects investment in
the home’s condition and appeal, which likely increases its market value.
BackyardSpace (quantitative): More outdoor space typically adds value, especially
in suburban and family-oriented markets.
Fireplace (categorical): Fireplaces are often viewed as a luxury or desirable feature,
potentially adding to a home’s price.
SchoolRating (quantitative): Higher-rated schools are associated with desirable
neighborhoods, often correlating with higher home prices.
These variables were selected to ensure a comprehensive model that includes both
structural and locational features commonly used in real-world housing valuation models.
C2. Descriptive Statistics
To better understand the variables in the dataset, descriptive statistics were computed for
all selected features. For continuous variables (SquareFootage, RenovationQuality,
BackyardSpace, SchoolRating, and Price), the analysis includes the mean, standard
deviation, and the five-number summary: minimum, first quartile (Q1), median, third
quartile (Q3), and maximum.
"In addition to summary statistics, frequency distributions were calculated for the
categorical variables. For example, 'Garage' appeared in approximately 36% of the homes,
while 64% lacked one. Similarly, about 40% of homes had a 'Fireplace'. The target variable,
'IsLuxury', showed a roughly 50/50 distribution, with 3,500 non-luxury and 3,500 luxury
homes out of 7,000 total. These distributions help confirm variable balance and guide
modeling decisions."
, (Screenshots of the descriptive statistics outputs for each variable are included in the
appendix section of the report).
C3. Visualizations
Univariate visualizations were constructed to explore the distribution of each variable
individually. For continuous variables like Price and SquareFootage, histograms were used.
The histogram of Price showed a right-skewed distribution, with a high frequency of homes
priced between $200,000 and $300,000. The histogram for SquareFootage revealed a
normal-like distribution with a few larger homes extending the upper range.
For categorical variables such as Garage and Fireplace, bar charts were used. These
visualizations illustrated the relative proportions of homes with and without each feature.
For instance, the bar chart for Garage showed a higher count of homes without garages,
supporting the descriptive statistics.
Bivariate visualizations were created to assess relationships between each independent
variable and the dependent variable (Price). For continuous predictor variables (e.g.,
SquareFootage, BackyardSpace, RenovationQuality, and SchoolRating), scatterplots were
used to visualize linear relationships. These scatterplots generally showed positive
associations—for instance, as SquareFootage increased, so did Price.
For categorical variables such as Garage and Fireplace, side-by-side boxplots were created
to compare the distribution of Price between the two categories (e.g., Garage = Yes vs. No).
The boxplot for Garage indicated that homes with garages had noticeably higher median
prices than those without, supporting its inclusion as a predictor.
Each visualization is explained with plain-English interpretations in the appendix, and
screenshots of these plots are included to support the visual analysis.
Section D: Data Analysis
D1. Data Splitting
The dataset was split into training (75%) and test (25%) sets using the scikit-learn
`train_test_split` function. Categorical variables were one-hot encoded where applicable,
and quantitative variables were used without scaling, following D600 guidelines.
D2. Model Optimization
The logistic regression model was built using the `statsmodels` package. Backward stepwise
elimination was used to remove variables with p-values above 0.05, one at a time. This
process ensures that only statistically significant predictors remain in the optimized model.
