The most common method is to check whether there are still problems in data processing, such as text conversion, extreme value removal, sorting and so on. , and then check whether descriptive statistics conform to common sense and get the right one.
Although this does not affect the meaning and direction too much, remember to add 1 if it is less than 1, because forgetting to add 1 has made the opposite result.
Look at the result after a basic operation. If there is no expected result, you can increase or decrease the control variables, fine-tune the model, or combine various methods (such as grouping and cross multiplication, the result is better).
. In short, as long as the theoretical derivation is reasonable, you can always get the desired results, but if you really can't get out, don't get stuck in a dead end or add stars yourself. When changing variables, whether they are all irrelevant or individual irrelevant, we must first process, log, standardize and so on.