Guys, anyone calculated the Hurst component in python as per the logic below:
https://blog.quantinsti.com/hurst-exponent/
online "hurst" library requires min of 100 time periods ?
Hi Sugat,
Yes, that's correct. The "hurst" library in Python typically requires a minimum of 100 data points to calculate the Hurst exponent accurately. This requirement is based on statistical considerations, as having more data points generally improves the accuracy of the estimation.
If you try to pass fewer than 100 data points to the 'compute_Hc' function of the "hurst" library, you will encounter the following ValueError:
This error indicates that the library expects the input series to have at least 100 data points. This requirement is imposed because estimating statistical properties, such as the Hurst exponent, becomes more reliable with larger sample sizes.
Therefore, if your dataset contains fewer than 100 data points, you'll need to increase the length of your dataset to use the "hurst" library in Python.
Let us know if you need further assistance!
Thanks Varun. With daily data, we need to have close to 4.5 months of data ( 22 trading days x 4.5 ) before we can calculate the Hurst exponent. 100 days is too long.
https://blog.quantinsti.com/hurst-exponent/ this methodology was implemented in the hurst library ?
Hi Sugat,
The need to use a minimum of 100 data points when calculating the Hurst exponent is primarily driven by statistical considerations and the nature of the analysis involved.
Estimating the Hurst exponent involves analyzing the long-term memory or self-similarity properties of a time series. To obtain reliable estimates of these properties, a sufficiently large number of data points is necessary to capture the underlying patterns in the data. With a larger sample size, the estimates become more stable and reliable.
The requirement for a minimum of 100 data points is aimed at ensuring the reliability, robustness, and generalizability of the Hurst exponent estimates obtained from the analysis of time series data.
Regarding your question about the methodology implemented in the "hurst" library, while the specific implementation details might differ, both the blog's approach and the "hurst" library aim to estimate the Hurst exponent using the R/S analysis method.
The main differences may lie in the handling of edge cases, optimizations, or specific parameter choices within the functions. Overall, they follow a similar methodology for estimating the Hurst exponent based on the rescaled range analysis.
If you want to explore the source code, you can find it here
I hope this helps!
Thanks Varun.