I want to model the difference in temperature between an electronic instrument surface and the ambient air. I used a linear multiple step-wise regression model with six variables as the potential predictors: net radiation, incoming shortwave (solar) radiation, incoming longwave radiation, horizontal wind speed, air temperature, water vapor density.
The significant curve-like trend in the scater plot of the modeled and measured temperature difference indicated that the linear model form was likely wrong.
But, if we ignore the interactions between the cilmatic variables, which variable should have a non-linear influence on the heating or cooling of the instrument surface compared with the ambient air? The wind speed? And also the water vapor density?
Dear Xingchang Wang,
Please, specify, the dependent variable of regression is Gaussian or not. If the variable is not Gaussian, then the application of the linear model is in question and the construction of nonlinear regression is required.
Dear U. Dreher, the ambient temperature is the ambient air temperature. Thanks!
- Badges
- Science topic