Scribd
Upload
5 views

Notes

Uploaded by

s22232956
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
5 views

Notes

Uploaded by

s22232956
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

A stationary process means the forecast will be more accurate because the data

stays within a consistent range.

If the data is non-stationary, it means that the mean or trend is changing over time,
making predictions less reliable.

The PDF (Probability Density Function) helps understand how likely different
outcomes are, especially for continuous variables.

In the Autoregressive (AR) model, we look at the relationship of a variable with its
past values.

Yt = 𝜇 + 𝜙 1 𝑌 𝑡 − 1 + 𝑢 𝑡 Y t=μ+ϕ 1Y t−1+u t
The term 𝜇 μ represents the constant mean. The error term 𝑢 𝑡 u tcaptures any
unexpected changes. Key points: The error has a zero mean, meaning it averages out
to zero. Constant variance means the variability of the error stays the same over
time. There is no autocorrelation, meaning today’s error isn’t linked to yesterday’s.
The error term behaves like white noise, meaning it's unpredictable. The GARCH
(Generalized Autoregressive Conditional Heteroskedasticity) model helps to
understand how variance (or volatility) changes over time.

𝜎2t=𝛼 0 + 𝛼 1 𝑢 𝑡 − 1 2 + 𝛽 1 𝜎 𝑡 − 1 2 σ t 2=α 0+α 1u t−1 2+β 1σ t−1 2

The model combines both autoregressive (impact from past errors) and moving
average (impact from past volatility). This model is great for explaining volatility
clustering, where periods of high volatility follow each other, and the same for low
volatility. The conditional variance depends on past periods. In the ARMA model
(Autoregressive Moving Average), we combine both autoregressive and moving
average components:

Yt=𝜇 + 𝜙 1 𝑌 𝑡 − 1 + 𝑢 𝑡 + 𝜃 1 𝑢 𝑡 − 1 Y t=μ+ϕ 1Y t−1+u t+θ 1u t−1

The AR term depends on past values ( 𝜙 1 𝑌 𝑡 − 1 ϕ 1Y t−1), while the MA term


depends on past errors ( 𝜃 1 𝑢 𝑡 − 1 θ 1u t−1). Volatility clustering means that
volatility tends to stick together: when there’s high volatility, it often stays high for a
while, and the same for low volatility.

The conditional variance adjusts based on past shocks or errors. To check for ARCH
effects (whether variance changes over time):

The null hypothesis 𝐻 0 H 0assumes there’s no ARCH effect, meaning variance is


constant. The alternative hypothesis 𝐻 1 H 1assumes there is an ARCH effect,
meaning variance is influenced by past periods. When testing for significance:

We use a critical value (1.96) at a 5% significance level to decide whether to reject


the null hypothesis. Small p-values mean we reject the null hypothesis. Volatility
forecasting is often done using ARCH/GARCH models, which predict future volatility
based on past patterns.

If variance is non-stationary, it indicates that volatility is shifting over time. Group


discussion on events that affected financial markets:

2008 Financial Crisis: Had a big impact on price levels and caused major volatility.
2000 Dot-com Crisis: Many tech companies were overvalued, leading to a crash.
2020 COVID-19 Pandemic: Global shock brought significant market volatility. 2022
Russia-Ukraine War: Drove up energy prices and caused inflation to spike.

You might also like