How do I decide which variables to remove from my curve fitting model?
Conditions
- Significance level is typically set to 0.05.
- The model includes multiple predictor variables with associated p-values.
Reasoning, step by step
- Identify the p-value for each predictor variable in the fitted model.
- Compare each p-value to the significance threshold (e.g., 0.05).
- If the p-value is greater than the threshold, conclude that the variable is not statistically significant.
- Remove the insignificant variable from the model to simplify it.
Example
The table displays the header 'p-value' and the value '0.32' for the 'Wind Shear' predictor. The speaker states, "its p-value is far above 0.05. So we can say with certainty that wind shear does not have a statistical significance on the power output, and we can axe it from our model."
Common misconceptions
- Removing variables based solely on visual inspection of their curves without checking statistical significance.
- Assuming a p-value slightly above 0.05 always justifies removal without considering the specific context or required precision.
Watch the explanation
Connected concepts
Explore next
Related questions
In the fitted wind-turbine model shown in the clip, wind speed has a much larger effect on predicted power output than wind direction, air density, turbulence, or wind shear. This is evidenced by the steep S-shaped curve for wind speed in the Plot Slice window, whereas the curves for the other factors are nearly horizontal, indicating very small impacts.
Conditions: The analysis is based on the specific dataset and fitted model shown in the video segment.; A nonlinear model has been fitted using wind speed plus four other factors.
The Plot Slice window displays separate panels for each predictor variable in a fitted multi-factor model. To compare impacts, observe the steepness of the curve in each panel.
Conditions: A fitted model with multiple predictors is available.; The software can generate partial-effect style plots for each predictor.
A model attempts to fit the overall trend of the data. When an outlier is present, the fitting algorithm tries to minimize the error across all points, including the extreme one.
Conditions: The data can be represented as a two-dimensional scatter plot.; The model attempts to fit the overall trend.; There exist observations clearly distant from the main cluster.
The nonlinear model combines a sigmoidal dependence on wind speed with assumed linear contributions from additional wind-related factors. Mathematically, it is expressed as , where represents wind speed, and through represent other factors like wind direction, air density, turbulence, and wind shear.
Conditions: Multiple predictor factors are to be modeled together.; Wind speed is known to have a sigmoidal relation to power output.; Other factors are assumed linear for simplicity.
Answers are generated from source material and independently checked. Consult the original video or creator if something is unclear.