Dear all,
I want to know about the basic things that we should keep in mind for DATA before proceeding to any model.
Like if we are using stock returns Time series as DATA then it may be possible that DATA have anomalies. Like outliers, Trend, stationarity, normality etc. that we have to keep in mind and provide some treatment to these problems before applying to any DATA model.
1.) I want to ask what are other things beyond mentioned above that we have to consider????
2.) How to detect these problems and provide treatment to these problems of data before proceeding to any model????
Because it is often seen that new researchers take data as it is without treatment of such problems and apply models directly and get struck after that at Vague Results.
Software Machine has no fault.
It becomes the situation GARBAGE in GARBAGE out.
Thanks and regards...
I think it depends on the field of research. There is not any specific rule to follow, all will depend on what you want to do with your data and where the data is coming from.
Someone would say that old school approach is to go with
the Box-Jenkins Method for Time Series . Here is nice de
https://www.google.ba/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&uact=8&ved=0ahUKEwjJ_ZTG1YvYAhUIzRQKHevbACEQFggyMAE&url=https%3A%2F%2Fmachinelearningmastery.com%2Fgentle-introduction-box-jenkins-method-time-series-forecasting%2F&usg=AOvVaw1dhFn6ey9dQqV534L35bN-
Now, I assume you want to find out whay there is so meny result that do not go on the line with that method.
Well, it seems like now there is something like whole other "club" that is on the oposite side.
For high frequancy data, I see no reason why not to go with procedure that has been sugested by Johansen "club" that is strick regarding applaying full procedure.
In macro data you have different story. Often there are not so many good quality data as it is the case with the data from financijal markets. Some methodes, like VAR framework allow more freedom in that sense than it is the case with the clasical time series approach.
- Badges
- Science topic
- Similar topics
- Psychology
- Cognitive Science
- Mind