Linear Regression Model Techniques with Python, NumPy, pandas and Seaborn

[@brentnunn2595]

Thank you Matt. You explain things so well. I've watched several of your videos and intend to watch more.

[@MattMacarty]

Thanks very much. Glad it's helping.

[@PaoloCondo]

Thank you for your work and the video!

[@MattMacarty]

Thanks. Glad it helped

[@tialynn1111]

Thank you! Is this also good for predicting a range? Say 30 periods out?

[@MattMacarty]

Glad it helped. It's tough to predict out that far, but yes it might give you something better than just guessing.

[@tialynn1111]

@@MattMacarty thank you Matt!

[@get_busy_living]

Thank you for this. I was looking for a way to code a trend line. With many traders it becomes subjective. This makes it more precise.

[@MattMacarty]

Thanks. Glad it helped

[@aimenslamat1264]

@@MattMacarty hi bro thank u so much for the video its help a lot , but could u tell me why we use the np log diff ?

[@MattMacarty]

Great. Glad it helped.

[@MattMacarty]

Glad it helped. Yes you can copy tables. see this video: th-cam.com/video/68XMlO9fciA/w-d-xo.html

[@MattMacarty]

@@aimenslamat1264 The log change is referred to as the instantaneous rate of return and is commonly used to price securities, particularly derivatives

[@iamekid8299]

Hi there, firstly great video! How would it be possible to add another variable to the trend line? So essentially it would be multiple linear regression

[@MattMacarty]

Thanks. Generally two explanatory variables would be graphed as a plane.

[@alvarezhenao]

Hello Matt, first of all, thank you for your work. I have tried to reproduce this exercise and when executing the line
'data = pdr.get_data_yahoo(stocks, start)'
I get the following error message:
'TypeError: string indices must be integers, not 'str'.
I have tried to solve it in different ways, but nothing, it always gives me that message. I downloaded your code in case there was something written incorrectly, and it gives me the same error. Any suggestions?

[@MattMacarty]

Thanks. The problem is the pandas_datareader has not kept up with changes to the yahoo finance API. The workaround is to install yfinance (pip install yfinance). Then it would be yfinance.download(stocks, start)

[@vimalk78]

@@MattMacarty i also faced same issue.. can i send a github PR to update the notebook if already not updated?

[@MattMacarty]

@@vimalk78 Sure I will post updated code using yfinance.

[@amanrai5285]

Thank you sir for explaing the code in the detail

[@amanrai5285]

explaining*

[@MattMacarty]

Glad it helped

[@Joe-xo2yd]

Matt, if I wanted to take the first derivative of the linear regression function, how could I execute that? is there a certain package?

[@MattMacarty]

Sure, you can use NumPy, SciPy, scikit-learn. Probably a number of others but that's what comes to mind.

[@nicholasrackard-hilt1619]

Thank you for the video!! How would I add another more trend lines to subset set specific times or dates? I am trying to plot my X and Y for 9 am 10am 12pm and have multiple trend lines. Hope that makes sense, thank you in advance!

[@MattMacarty]

You mean for different time periods, say an hour for each? You should be able to subset the data and then run a regression on each subset.

[@philtoa334]

Very good Thank you.

[@MattMacarty]

Glad it helped

[@Shinykip]

Could you possibly provide a link to the data set you actually used in this tutorial? I would like to follow it locally on my PC as I go through this. Thank you.

[@MattMacarty]

Hi the data was downloaded from Yahoo finance. Here is a link to the notebook: github.com/mjmacarty/alphavantage

[@satonaka123]

Thank you Matt. Would like to know what is different btw "np.log().diff()" and "pct_change().cumsum()", I've tried those two syntax then got a different result. Could you pls solve this question? Thanks again Matt.

[@MattMacarty]

pct_change is E/B - 1, np.log is LN(E/B). LN is a continuously compounding rate.

[@eliotharreau7627]

Very Nice, Hello Matt, in the section "trend as a technical indicateur", how to put the value of the trend line in a dataframe to detect later when it cross the curve ? Thank you. 👍👍

[@MattMacarty]

You need the slope and intercept coefficient and then you can create a column with the trend line

[@eliotharreau7627]

@@MattMacarty Yess my friend it s exactly what i want, maybe if you have a small example ? Thnx.

[@mamunmd.abdullahall5122]

import pandas_datareader as pdr is not working , what shoul di do

[@MattMacarty]

This probably means it is not installed. From the command line run pip install pandas-datareader

[@marcus.guitarist]

@@MattMacarty looks like Yahoo might have changed something

[@tonyhathuc]

Hi I am running into this error:
No data fetched using 'YahooDailyReader'

[@MattMacarty]

Can you send me the line of code that caused the error?

[@anthonychung2907]

what to do if the x-axis is time series ?

[@MattMacarty]

Probably the easiest thing to do is make time the independent variable.

[@jamesnelligan660]

i get error: "ModuleNotFoundError: No module named 'pandas_datareader'"

[@MattMacarty]

Hi this means you need to install it. Run !pip install pandas-datareader in a code cell in your notebook

[@danieleliahushapiro4280]

Correlation != Causality !!!

[@MattMacarty]

Exactly

[@fuhat1040]

i got two points

[@MattMacarty]

Hi. You can download the notebook I used here: alphabench.com/data/python-linear-regression.html

[@fuhat1040]

@@MattMacarty yeah i still got two points for spy and Google might be due to a crash

[@fuhat1040]

do you want to look at my notebook when I finish it.

[@fuhat1040]

changed the symbols still got two points using goog