2010 - 2016 greater than 2.5

How does today's yield curve compare to those in the run-up to the 2000 & 2008 market collapses?

Not only have bond prices been rising forever, but they paid out more than stocks. Might that all be about to change?

A sliding window analysis using the S&P500 answers the question of how well a set of investments and withdrawals would have performed in each one of the N-year-wide windows from 1950. So a set of 30-year windows looks at 1950-1979, 1951-1980, etc. up the the partial period ending with the current year. This package takes two vectors: investments by year and withdrawals by year plus the width of the window and produces graphs and statistics showing the results in each window. The investment and withdrawal vectors are annual (and are assumed to be made on the first trading day of the year) but the analysis is done on the daily close of the S&P 500 Total Return Index with dividends reinvested.

An important aspect of Health Savings Accounts is the investment return that can be expected for the average investor. This analysis creates medical-expense and investment-propensity distributions over an assumed 30-years of investing and then uses stock-market data since 1950 to show both the range of outcomes and the most-likely ones.

The r tutorial for the kaggle contest

As a project for the course I attempted to predict the survivors of the Titanic

Detailed walkthrough of possibilities

Trees

Use the models built on the training data on the test data

Singular Value Decomposition (SVD)

Exploratory Data Analysis

Data from Samsung Galaxy 3 smart phone accelerometers was downloaded and subsetted

Two models were built: a simple quadratic using only the FICO varaible and an elaborate multiple linear regression model which started with all the variables and reduced them to maximize the model parameters. Run against the test half of the dataset, the elaborate model produced a Mean Squared Error 57% smaller than the simple model.

In the previous step, several variables show certain peculiarities. In this step I looked to see if transforms could improve their predictive power.

Each variable in the cleaned-up dataset was examined for any peculiar features and to gain a deeper understanding of the data and possible interactions.

The loan data file contained data in formats that were incompatible with regression analysis. In addition, certain factors were missing and had to be added for downstream use.

A demonstration of an R function to allow you to source R scripts from Dropbox files.