如何绘制标准差误差线？

huangapple
go评论120阅读模式
英文:

how do I plot a standard deviation error line?

问题

使用scipy，我可以绘制标准差线。如何在该线的上方复制一个95%的线，以及在该线的下方复制另一个95%的线。

请查看下面的两张图片，以了解我的意思：

如何绘制标准差误差线？

如何绘制标准差误差线？

我的代码：

  1. import datetime as date
  2. import matplotlib.pyplot as plt
  3. %matplotlib inline
  4. import datetime
  5. from scipy import stats
  7. Ins_Name = "EURUSD=X"
  8. # Ins_Name = "AAPL"
  9. df = yf.download(Ins_Name,'2019-05-01','2020-01-03')
  11. df.reset_index(inplace=True)
  12. df['date_ordinal'] = pd.to_datetime(df['Date']).apply(lambda date: date.toordinal())
  13. DateVariance = [datetime.date(2019, 5, 1), datetime.date(2020, 1, 3)]
  15. x_reg = df.date_ordinal
  16. y_reg = df.Close
  17. slope, intercept, r_value, p_value, std_err = stats.linregress(x_reg, y_reg)
  18. print("slope: %f    intercept: %f     STD Error: %f" % (slope, intercept, std_err))
  20. sns.set()
  21. # plt.figure(figsize=(26, 10))
  22. fig, ax = plt.subplots(figsize=(15,7))
  23. ax = plt.plot(x_reg,intercept + slope*x_reg, color='b')
  24. # ax = sns.lmplot('date_ordinal', 'Close', data=df, fit_reg=True, aspect=2, ) #Scatter PLot
  25. ax = sns.regplot(data=df,x='date_ordinal',y=df.Close,ci=1,fit_reg=False,scatter_kws={"color": "red"}, line_kws={"color": "black"}, marker='x') #scatterplot
  26. # sns.jointplot('date_ordinal', df.Close, data=df, kind="reg",ylim=[1.089,1.15],xlim=DateVariance, height=12,color='red',scatter_kws={"color": "red"}, line_kws={"color": "black"})
  28. ax.set_xlabel('Interval', fontsize=25)
  29. ax.set_xlim(DateVariance)
  30. ax.set_ylabel('Price', fontsize=25)
  31. ax.set_title('Mean reversion of ' + Ins_Name + ' Close Prices',fontsize= 45,color='black')
  32. new_labels = [datetime.date.fromordinal(int(item)) for item in ax.get_xticks()]
  33. ax.set_xticklabels(new_labels)

希望这能帮助你。

英文:

Using scipy I can plot the standard deviation line. How do I duplicate that same line 95% above and another one 95% below.

See the two images below of what I mean:

如何绘制标准差误差线？

如何绘制标准差误差线？

My code:

  1. import datetime as date
  2. import matplotlib.pyplot as plt
  3. %matplotlib inline
  4. import datetime
  5. from scipy import stats
  7. Ins_Name = "EURUSD=X"
  8. #Ins_Name = "AAPL"
  9. df = yf.download(Ins_Name,'2019-05-01','2020-01-03')
  11. df.reset_index(inplace=True)
  12. df['date_ordinal'] = pd.to_datetime(df['Date']).apply(lambda date: date.toordinal())
  13. DateVariance = [datetime.date(2019, 5, 1), datetime.date(2020, 1, 3)]
  15. x_reg = df.date_ordinal
  16. y_reg = df.Close
  17. slope, intercept, r_value, p_value, std_err = stats.linregress(x_reg, y_reg)
  18. print("slope: %f    intercept: %f     STD Error: %f" % (slope, intercept, std_err))
  20. sns.set()
  21. #plt.figure(figsize=(26, 10))
  22. fig, ax = plt.subplots(figsize=(15,7))
  23. ax = plt.plot(x_reg,intercept + slope*x_reg, color='b')
  24. #ax = sns.lmplot('date_ordinal', 'Close', data=df, fit_reg=True, aspect=2, ) #Scatter PLot
  25. ax = sns.regplot(data=df,x='date_ordinal',y=df.Close,ci=1,fit_reg=False,scatter_kws={"color": "red"}, line_kws={"color": "black"}, marker='x') #scatterplot
  26. #sns.jointplot('date_ordinal', df.Close, data=df, kind="reg",ylim=[1.089,1.15],xlim=DateVariance, height=12,color='red',scatter_kws={"color": "red"}, line_kws={"color": "black"})
  28. ax.set_xlabel('Interval', fontsize=25)
  29. ax.set_xlim(DateVariance)
  30. ax.set_ylabel('Price', fontsize=25)
  31. ax.set_title('Mean reversion of ' + Ins_Name + ' Close Prices',fontsize= 45,color='black')
  32. new_labels = [datetime.date.fromordinal(int(item)) for item in ax.get_xticks()]
  33. ax.set_xticklabels(new_labels)

答案1

得分: 0

我认为函数 stats.linregress() 并不是你想要的。它只返回斜率的不确定性，而不是 y 截距，如果我理解你的问题正确的话，你想要的是 y 截距。

我建议使用 scipy.optimize.curve_fit()，像下面这样：

  1. import datetime as date
  2. import matplotlib.pyplot as plt
  3. import datetime
  4. from scipy.optimize import curve_fit
  5. import yfinance as yf
  6. import pandas as pd
  7. import numpy as np
  8. import seaborn as sns
  10. def fline(x, *par):
  11.     return par[0] + par[1]*x
  13. Ins_Name = "EURUSD=X"
  14. #Ins_Name = "AAPL"
  15. df = yf.download(Ins_Name,'2019-05-01','2020-01-03')
  17. df.reset_index(inplace=True)
  18. df['date_ordinal'] = pd.to_datetime(df['Date']).apply(lambda date: date.toordinal())
  19. DateVariance = [datetime.date(2019, 5, 1), datetime.date(2020, 1, 3)]
  21. x_reg = df.date_ordinal
  22. x_trans = x_reg - x_reg[0]
  24. y_reg = df.Close
  25. pi = [1.1, -0.0001] # 初始猜测值，可以使用 linregress 获取这些值
  26. # 在这里执行曲线拟合
  27. popt, cov = curve_fit(fline, x_trans, y_reg, p0=pi) # 返回值和协方差（误差）矩阵
  28. yInterceptSigma = np.sqrt(cov[0,0])
  29. print(x_reg)
  30. print(x_reg - x_reg[0])
  31. print(popt)
  32. print(cov)
  34. sns.set()
  35. #plt.figure(figsize=(26, 10))
  36. fig, ax = plt.subplots(figsize=(15,7))
  37. ax = plt.plot(x_reg, fline(x_trans, *popt), color='b')
  38. # +/- 2sigma gives ~95% CI
  39. plt.plot(x_reg, fline(x_trans, *[popt[0] + 2 * yInterceptSigma, popt[1]]), '--b') # +2sigma on the yintercept
  40. plt.plot(x_reg, fline(x_trans, *[popt[0] - 2 * yInterceptSigma, popt[1]]), '--b') # -2sigma on teh yintercept
  41. #ax = sns.lmplot('date_ordinal', 'Close', data=df, fit_reg=True, aspect=2, ) #Scatter PLot
  42. ax = sns.regplot(data=df,x='date_ordinal',y=df.Close,ci=1,fit_reg=False,scatter_kws={"color": "red"}, line_kws={"color": "black"}, marker='x') #scatterplot
  43. #sns.jointplot('date_ordinal', df.Close, data=df, kind="reg",ylim=[1.089,1.15],xlim=DateVariance, height=12,color='red',scatter_kws={"color": "red"}, line_kws={"color": "black"})
  45. ax.set_xlabel('Interval', fontsize=25)
  46. ax.set_xlim(DateVariance)
  47. ax.set_ylabel('Price', fontsize=25)
  48. ax.set_title('Mean reversion of ' + Ins_Name + ' Close Prices',fontsize= 45,color='black')
  49. new_labels = [datetime.date.fromordinal(int(item)) for item in ax.get_xticks()]
  50. ax.set_xticklabels(new_labels)
  52. plt.show()

这将生成以下图表：如何绘制标准差误差线？

英文:

I think the function stats.linregress() is not what you want here. It only returns the uncertainty on the slope and not the y-intercept, which, if I understand your question, is what you want.

I would use scipy.optimize.curve_fit() like below:

  1. import datetime as date
  2. import matplotlib.pyplot as plt
  3. import datetime
  4. from scipy.optimize import curve_fit
  5. import yfinance as yf
  6. import pandas as pd
  7. import numpy as np
  8. import seaborn as sns
  10. def fline(x, *par):
  11.     return par[0] + par[1]*x
  13. Ins_Name = "EURUSD=X"
  14. #Ins_Name = "AAPL"
  15. df = yf.download(Ins_Name,'2019-05-01','2020-01-03')
  17. df.reset_index(inplace=True)
  18. df['date_ordinal'] = pd.to_datetime(df['Date']).apply(lambda date: date.toordinal())
  19. DateVariance = [datetime.date(2019, 5, 1), datetime.date(2020, 1, 3)]
  21. x_reg = df.date_ordinal
  22. x_trans = x_reg - x_reg[0]
  24. y_reg = df.Close
  25. pi = [1.1, -0.0001] # initial guess. Could use linregress to get these values
  26. # Perform curve fitting here
  27. popt, cov = curve_fit(fline, x_trans, y_reg, p0=pi) # returns the values and the covariance (error) matrix
  28. yInterceptSigma = np.sqrt(cov[0,0])
  29. print(x_reg)
  30. print(x_reg - x_reg[0])
  31. print(popt)
  32. print(cov)
  34. sns.set()
  35. #plt.figure(figsize=(26, 10))
  36. fig, ax = plt.subplots(figsize=(15,7))
  37. ax = plt.plot(x_reg, fline(x_trans, *popt), color='b')
  38. # +/- 2sigma gives ~95% CI
  39. plt.plot(x_reg, fline(x_trans, *[popt[0] + 2 * yInterceptSigma, popt[1]]), '--b') # +2sigma on the yintercept
  40. plt.plot(x_reg, fline(x_trans, *[popt[0] - 2 * yInterceptSigma, popt[1]]), '--b') # -2sigma on teh yintercept
  41. #ax = sns.lmplot('date_ordinal', 'Close', data=df, fit_reg=True, aspect=2, ) #Scatter PLot
  42. ax = sns.regplot(data=df,x='date_ordinal',y=df.Close,ci=1,fit_reg=False,scatter_kws={"color": "red"}, line_kws={"color": "black"}, marker='x') #scatterplot
  43. #sns.jointplot('date_ordinal', df.Close, data=df, kind="reg",ylim=[1.089,1.15],xlim=DateVariance, height=12,color='red',scatter_kws={"color": "red"}, line_kws={"color": "black"})
  45. ax.set_xlabel('Interval', fontsize=25)
  46. ax.set_xlim(DateVariance)
  47. ax.set_ylabel('Price', fontsize=25)
  48. ax.set_title('Mean reversion of ' + Ins_Name + ' Close Prices',fontsize= 45,color='black')
  49. new_labels = [datetime.date.fromordinal(int(item)) for item in ax.get_xticks()]
  50. ax.set_xticklabels(new_labels)
  52. plt.show()

Which gives the following plot:
如何绘制标准差误差线？

huangapple
  • 本文由 发表于 2020年1月6日 20:36:03
  • 转载请务必保留本文链接：https://go.coder-hub.com/59612243.html
  • matplotlib
  • python
  • scipy
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定