This document analyzes return time series data for the S&P 500 and Russell indexes at daily, weekly, monthly, and yearly frequencies. Key analyses include:
- Calculating descriptive statistics for each return series.
- Plotting distributions and finding they appear normally distributed.
- Conducting Shapiro-Wilk tests which reject the null hypothesis of normality for all series.
- Finding best fit t-distributions provide better fits to the data than normal distributions.
- Concluding returns are better modeled by t-distributions and frequency decreases from daily to yearly with little difference between S&P and Russell indexes.
This document analyzes return time series data for the S&P 500 and Russell indexes at daily, weekly, monthly, and yearly frequencies. Key analyses include:
- Calculating descriptive statistics for each return series.
- Plotting distributions and finding they appear normally distributed.
- Conducting Shapiro-Wilk tests which reject the null hypothesis of normality for all series.
- Finding best fit t-distributions provide better fits to the data than normal distributions.
- Concluding returns are better modeled by t-distributions and frequency decreases from daily to yearly with little difference between S&P and Russell indexes.
This document analyzes return time series data for the S&P 500 and Russell indexes at daily, weekly, monthly, and yearly frequencies. Key analyses include:
- Calculating descriptive statistics for each return series.
- Plotting distributions and finding they appear normally distributed.
- Conducting Shapiro-Wilk tests which reject the null hypothesis of normality for all series.
- Finding best fit t-distributions provide better fits to the data than normal distributions.
- Concluding returns are better modeled by t-distributions and frequency decreases from daily to yearly with little difference between S&P and Russell indexes.
This document analyzes return time series data for the S&P 500 and Russell indexes at daily, weekly, monthly, and yearly frequencies. Key analyses include:
- Calculating descriptive statistics for each return series.
- Plotting distributions and finding they appear normally distributed.
- Conducting Shapiro-Wilk tests which reject the null hypothesis of normality for all series.
- Finding best fit t-distributions provide better fits to the data than normal distributions.
- Concluding returns are better modeled by t-distributions and frequency decreases from daily to yearly with little difference between S&P and Russell indexes.
Copyright:
Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online from Scribd
Download as pdf or txt
You are on page 1/ 12
Anurag Ramachandra Assignment 1 1.
Create daily, weekly, monthly, and yearly return and log return time series (16) and plot them.
Anurag Ramachandra Assignment 1
Anurag Ramachandra Assignment 1
Anurag Ramachandra Assignment 1
Anurag Ramachandra Assignment 1
Anurag Ramachandra Assignment 1 2. Compute sample mean, sample stdev, skewness and kurtosis for each data set S & P Daily Mean: "-0.000241030663573117" Std Dev: "0.00981174902707828" Skew: "0.0147062329389353" Kurtosis: "0.00396382214435235" S & P Weekly Mean: "-0.00116909104789517" Std Dev: "0.0209421768332882" Skew: "0.0170214036808351" Kurtosis: "0.00443177586216148" S & P Monthly Mean: "-0.00512412679711084" Std Dev: "0.0424254196320831" Skew: "0.0393696930135791" Kurtosis: "0.0118027056967112" S & P Yearly Mean: "-0.0585697805259767" Std Dev: "0.159502616998138" Skew: "0.296402862894988" Kurtosis: "0.213161935219314" S & P Log Mean: "-0.000288826508699675" Std Dev: "0.00975700886392262" Skew: "0.0100438438835809" Kurtosis: "0.00291879484672612" Russell Daily Mean: "-0.000192609487307321" Std Dev: "0.0134004258506673" Skew: "0.0108897557789958"
Anurag Ramachandra Assignment 1 Kurtosis: "0.00229874493082872" Russell Weekly Mean: "-0.000938229709841227" Std Dev: "0.0291990074165653" Skew: "0.0418151720597322" Kurtosis: "0.0107624706078601" Russell Monthly Mean: "-0.00409374919440765" Std Dev: "0.061470273585821" Skew: "0.120713217648036" Kurtosis: "0.0480787014901302" Russell Yearly Mean: "-0.0636014834445323" Std Dev: "0.198459306802261" Skew: "0.306936363076424" Kurtosis: "0.240969877721371" Russell Log Mean: "-0.000281881707802958" Std Dev: "0.0133430518679411" Skew: "0.00798070923107131" Kurtosis: "0.00210260485189475" 2. Plot density function and cumulative distribution function for the empirical distributions together with normal distribution with the same mean and variance. Do the returns, log returns look normally distributed?
S&P 500
Anurag Ramachandra Assignment 1
Anurag Ramachandra Assignment 1
Russel
Anurag Ramachandra Assignment 1
They all appear to be normally distributed.
Anurag Ramachandra Assignment 1 4. Test for normality using the Shapiro-Wilk test? What is the p-value? Can you reject the null hypothesis of a normal distribution at 0.01 in each case?
S & P Daily W = 0.9358, p-value < 2.2e-16 S & P Weekly W = 0.9459, p-value < 2.2e-16 S & P Monthly W = 0.9597, p-value = 1.075e-13 S & P Yearly W = 0.8555, p-value = 2.421e-06 S & P Log W = 0.9382, p-value < 2.2e-16 Russell Daily W = 0.9181, p-value < 2.2e-16 Russell Weekly W = 0.9003, p-value < 2.2e-16 Russell Monthly W = 0.884, p-value = 9.027e-15 Russell Yearly W = 0.8764, p-value = 0.00483 Russell Log W = 0.924, p-value < 2.2e-16 5. Create several t-plots (student t distribution probability plots) of the returns and log returns using a number of choice of the degrees-of-freedom parameter (df). What value of df gives a plot that is as linear as possible? Degrees of Freedom
S&P S & P Daily: 3 S & P Weekly: 4/5 S & P Monthly: ~30 S & P Yearly: ~100
Anurag Ramachandra Assignment 1 Log S & P Daily: 3 S & P Weekly: 4/5 S & P Monthly: ~20 S & P Yearly: ~100 Russell Russell Daily: 3 Russell Weekly: 4 Russell Monthly: ~30 Russell Yearly: ~100 Log Russell Daily: 3 Russell Weekly: 4 Russell Monthly: ~10/20 Russell Yearly: ~100 6. Write your conclusions for each data set.
Given how few data points we have and how the Shapiro-Wilk test returns very small p-values for each of these data sets, we can clearly reject the null hypothesis of normality. The heavy tails make it far more likely that the data-series for all time frequencies comes from t-distributions. Best fit t-distributions were obtained for all data-series and what can be concluded is that the frequency decreases from daily to yearly. Interesting to note is that all (inflation adjusted) price data-series have a slightly negative mean. There does not seem to be much difference between the distributions for the small cap (Russel) versus the large cap (S&P), compared at the same time intervals (daily, weekly etc.).
