Time series Flashcards
What are different types of temporal variation?
Step, cyclic, episode, trend, noise
What does r2-value say?
It says how large the of y is that is explained by x
Which criteria need to be fulfilled in a linear regression?
- Constant variance
- Linear model is correct
- Independ residuals (no autocorrelation)
- Normally distrubted residuals
What relaxtions can be made compared to a linear regression when a non-parametric method like Mann-Kendall is used?
- The trend only needs to be monotonic
- Variance does not need to be constant
- Residuals do not need to be normally distrubeted
What is Thiels slope?
It is the median of all pairwise slopes
What is the benefit of the seasonal Kendall?
The autocorrelation is handeled
What are the main goals of non-parametric regression?
- To describe and visualize the data
- estimate in which way the response variable depends on
some predictors without specifying the functional form
Describe the difference between parametric and non-parametric models?
In a paramteric model we always have tp specify the functional form and is useful if we know that something should follow a certain relationship. Have harder criteria.
In a non-parametric fitting the form is completly free. Less criteria need to be fulfilled.
What is cross-validation?
To see how good a certian model is the data is dived into a set of training data and new data that is used how good the model is in predicting new data. By this different degrees of smoothing can be tried.
What is GAM?
Is a model in which different splines are combined to a prediction curve. Is a non-parametric method. Can consist of different types of functions that are combined
What are effective degrees of freedom?
In GAM you cannot determine DF. You therefore get the EDF, which gives an indication of how complex the relationship is. The closer to 1 the value is the more linear is the model.
