Timeline for Regression coefficient has negative symbol but positive from the raw plot
Current License: CC BY-SA 4.0
6 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 27, 2018 at 21:21 | comment | added | Ben | It sounds like you are interested in modelling the causal effects of these variables. This will require you to read up on controlled experimentation and regression. If you wish to make causal inferences (i.e., interpret your parameters as causal effects) then this generally requires you to engage in some kind of controlled trial to avoid problems with confounding variables. Without this you can still interpret your coefficients predictively, in the sense that they are simply rates-of-change of the conditional expectation of the response variable. | |
| Nov 27, 2018 at 16:29 | comment | added | Nick Cox | "casual" to "causal" I think. | |
| Nov 27, 2018 at 1:45 | vote | accept | 89_Simple | ||
| Nov 27, 2018 at 0:18 | comment | added | 89_Simple | Thanks Ben for your time and explanation. Would it be possible for you to add addition explanations. My x1 and x2 are variables that measure the water availability to crops so as x1 or x2 increases (better water availability), the yield should go up as well (i.e. a positive correlation of x1 and x2 with yield which the univariate plots show. Does this result mean that I cannot use this model for any prediction since the coefficient of x1 is wrong (negative indicting yield goes down with increasing x1) or does it mean that interpreting the reg coefficients as it is not practical in this case? | |
| Nov 26, 2018 at 23:56 | history | edited | Ben | CC BY-SA 4.0 |
added 1604 characters in body
|
| Nov 26, 2018 at 23:23 | history | answered | Ben | CC BY-SA 4.0 |