Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 2

Points to be remembered.

 We have to show how customer count has been decreased and we have to show the
decreasing trend.
 Total Credit limit is related to customer count. Decrease with customer count.
 Average open to buy is also decreasing
 Total transaction count also decreasing.
 Average utilization ratio should be constant.

Flow of Conduct

 Check whether the model is time series or not , plot graph


 Clean the data by reducing the outliers
 Check for the stationary using ADF. Run the regression model customer account wrt the
variables.(Can use lagged methods to achieve stationarity)
 See if there is any trend in the graphs of each variable, use time dummy variable,
differencing method, decomposing method. Check all three processes
 Check between the initial models using Anova.
 Check for IID for the errors.(Errors should be constant).
 If possible try to check for random walk.
 Check for serial correlation with help of ACF and PACF and try to solve the issue of
correlation using differencing method.
 Use the regression of residuals on the independent variables and lagged variables.
 If the serial correlation is left use FGLS to remove serial correlation.
 Try to find PDQ using auto dot ARIMA.
 Run the ARIMA model
 Check for heteroskedascticity and also exogenity.
To Check for Stationarity

How to test if a time series is stationary?


Use Augmented Dickey-Fuller Test (adf test). A p-Value of less than 0.05 in adf.test()
indicates that it is stationary.

library(tseries)
adf.test(tsData) # p-value < 0.05 indicates the TS is stationary
kpss.test(tsData)

2. Augmented Dickey-Fuller (ADF) t-test Small p-values suggest that the data


is stationary and doesn’t need to be differenced stationarity.

adf_test <- adf.test(quandldata[,1],alternative = 'stationary')


print(adf_test)

Run Linear Models.

linearMod <- lm(dist ~ speed, data=cars) # build linear regression model on full data
print(linearMod)

You might also like