- Residual fitted
- normal Q
- scale location
- residual leverage graph.
Not able to interpret what it means.
Asked
Active
Viewed 58 times
0
Ferdi
- 4,882
- 7
- 42
- 62
-
Welcome to StackOverflow. Please take the [tour](http://stackoverflow.com/tour) have a look around, and read through the [HELP center](http://stackoverflow.com/help), then read [How to Ask Question](http://stackoverflow.com/help/how-to-ask), [What types of questions should I avoid asking?](http://stackoverflow.com/help/dont-ask) and provide a [MCVE : Minimal, Complete, and Verifiable Example](http://stackoverflow.com/help/mcve). If people around can easily read and understand what you mean, or what the problem is, they'll be more likely willing to help :) – Dwhitz Jun 01 '18 at 06:34
-
From the first plot, it appears that you have curvature that you are not accounting for in the model. Perhaps a quadratic term for loan_id or status would be warranted? – mkt Jun 01 '18 at 10:48
-
What are the variables? Predicting one identifier from another is a puzzling problem, if that is what the names mean. Also, is either of the *id variables necessarily positive? (Status I guess wildly is an indicator or categorical variable.) – Nick Cox Jun 01 '18 at 12:34
1 Answers
1
The most obvious and important message is on the top left. When the predicted values (on the x axis) are larger than 40, the residuals are near zero, which is good. When they are smaller than 40, the residuals are systematically negative, meaning, these predictions are too high. Your simple linear model is not good for small values and you should seek to make your model a bit more complex, so smaller results become more accurate.
If you have an idea, what this means in terms of your question, try to model that idea. If not, try a quadratic term as @mkt said in the comments - or any other transformation (exponential, logarithmic, ...) that you have been taught.
