Hello everyone,
I've been struggling with this matter:
We have data on sugar cane production (soils, treatments, weather …). We want to understand in which conditions soil/plant nutrition will yield the best result. It is hard to infer causality as some treatments are often performed jointly and we are not able to isolate the partial effect of the applications of interest.
Any one has any leads on interesting statistical methods please ?
Thanks a lot
Hello Adrien,
The gold standard method, of course, to determining causal influence for a variable or variable set is to manipulate that variable (or set) on randomly assigned cases and collect results on the subsequent values of the outcome variable(s) of interest.
If that's not feasible, then what you can do is evaluate how much (if any) explanatory power the individual variable/s may have, being sure to include interaction terms among the variable set under consideration. Of course, there is always the possibility that you are omitting one or more important variables from the set being evaluated (or that there is something idiosyncratic about the responses observed from the data you did collect, which might not generalize, or could again represent an unmeasured influence). Multiple linear regression would be a plausible framework for initiating this type of analysis.
Good luck with your work.
Hello Adrien,
The gold standard method, of course, to determining causal influence for a variable or variable set is to manipulate that variable (or set) on randomly assigned cases and collect results on the subsequent values of the outcome variable(s) of interest.
If that's not feasible, then what you can do is evaluate how much (if any) explanatory power the individual variable/s may have, being sure to include interaction terms among the variable set under consideration. Of course, there is always the possibility that you are omitting one or more important variables from the set being evaluated (or that there is something idiosyncratic about the responses observed from the data you did collect, which might not generalize, or could again represent an unmeasured influence). Multiple linear regression would be a plausible framework for initiating this type of analysis.
Good luck with your work.
As @David Morse said Multiple Linear Regression would work. But two things need consideration. (1) Such data are collected using quasi-experimental research since experimental research is difficult to conduct as all the independent variable cannot be controlled in this case. (2) Note that independent variables may have synergistic effect, e.g. soil pH and fertilizer application.
You can share the condition in your experimental area to have a better understanding
Hi Adrien Kroese ,
As David Morse said, "Multiple linear regression would be a plausible framework for initiating this type of analysis" is a plausible framework for initiating this type of analysis.
However, be careful because non-experimental data could be unbalanced and biased.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics