Calibrating MA(10) model: Calibrated coefficients do not agree with the model in the text

General
1

Course Name: Financial Time Series Analysis for Trading, Section No: 15, Unit No: 7, Unit type: Notebook

Hello,

please find below an inconsistency between fitted parameters and the explanatory text in the Jupyter notebook.

ma.L1 0.21
ma.L2 -0.17
ma.L3 -0.08
ma.L4 0.04
ma.L5 -0.22
ma.L6 0.11
ma.L7 0.11
ma.L8 0.09
ma.L9 -0.26
ma.L10 0.34
sigma2 0.75
dtype: float64
From the output above, you can see that the fitted model is
๐‘€๐ด(10)=๐‘ฆ๐‘ก=โˆ’0.23+๐œ–๐‘ก+0.09โˆ—๐œ–๐‘กโˆ’1+ โ€ฆ +0.22โˆ—๐‘ฆ๐‘กโˆ’10

Kindly update the notebook.

Also, further down in the notebook, we find the following:
" From the third plot, you can see that there is autocorrelation between the residuals with its 3rd and 8th lagged value."
In the present form the PACF plot does not indicate any autocorrelation of the residuals within the 95% confidence interval [I assume from data_modules.tsa_functions_quantra import model_performance indeed used 95% confidence intervals?].

  1. Is the code from within data_modules.tsa_functions_quantra accessible from, via github, so one can see what it is actually doing?

Many thanks for your help.

2

Hi Benjamin, we are looking into your questions with regards to the material. Will get back to you shortly.

3

Thanks for pointing this out.

You are right about both issues. The MA(10) equation has been corrected to match the fitted coefficients and now properly uses error terms instead of a lagged term.

The residual PACF interpretation has also been updated โ€” it now reflects that no lags are statistically significant within the 95% confidence interval.

Regarding data_modules.tsa_functions_quantra, this is included in the course resources. You can access the implementation locally after downloading the course files, where model_performance is defined.

Appreciate your feedback.