Difference between LinearRegression and LinearRegressionSlope?

• 01/16/2021 at 9:00 PM #158093
  vsoler

  I am having trouble using the LinearRegressionSlope and LinearRegression indicators. The help texts do not provide enough information, so I decided to ask this forum.

  1. if I want to calculate a regression line using the last “N” daily bars, it has to be calculated on the last “N-1” daily returns. When using the LinearRegressionSlope function, do I have to use “N” or “N-1” as a parameter in brackets?  If I want to calculate on the last 100 bars, should I use LinearRegrassionSlope[100](close) or LinearRegressionSlope[99]
  2. the number returned by the LinearRegressionSlope function, is it given in % or in “$/day”?
  3. the number returned by the LinearRegression function, is it the intercept?  is it given in $?

  If someone could provide some light, it would be very helpful to me. I am unable to match the figures returned by these 2 function with those calculated on an excel spreadsheet.

  Thanks in advance

  

  vsoler

  Participant

  Topics: 4

  Replies: 2

  Been thanked: 0 times
  01/16/2021 at 9:54 PM #158095
  Vonasi

  LinearRegrassionSlope[100](close) will return the value calculated on the last 100 bars including the still currently forming bar.

  LinearRegrassionSlope[100](close[1]) will return the value calculated on the last 100 bars calculated at the close of the bar prior to the still currently forming bar. i.e yesterdays value at closing price on a daily chart.

  Where are you getting your Excel spreadsheet data from because if it is not identical to the data feed from your broker to the PRT platform then the calculated result is very unlikely to be identical?

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  01/17/2021 at 8:15 PM #158192
  vsoler

  Thank you very much for your quick reply,

  In PRT, I use the close adjusted by dividends and splits

  In Excel, I use Yahoo Finance historical data, of which I use the “adjusted-close” column.

  I think I found the mismatch, and everything is now OK.

  What happened was that with candlesticks, I used data for close to close[66]. However, for the LinearRegressionSlope function I have to use 67 candlesticks.

  Thank you very much

  

  vsoler

  Participant

  Topics: 4

  Replies: 2

  Been thanked: 0 times
  05/15/2022 at 1:00 AM #193194
  qigley

  Does anyone know what the figure returned by the Linear Regression Slope function means?

  It should mean rate of change of Y divided by rate of change of X. For the 6 candle range on stock I am using it on this would mean  about 20 price points over 6 candles, resulting in -3.33.

  However the figure it is giving me is  -0.058. I cant correlate that to the visible slope (white unbroken line going through the candles)

  

  

  qigley

  Participant

  Topics: 2

  Replies: 18

  Been thanked: 4 times
  05/15/2022 at 11:32 AM #193211
  JS

  The indicator calculates the slope of the most appropriate line using the least squares method.

  If you use this indicator with a period of 6, the calculation also includes the last (accrual) bar[0], so then the figure of -0.058 could be correct.

  

  JS

  Participant

  Topics: 26

  Replies: 1720

  Been thanked: 782 times
• Author
  Posts