16

Hi i have been learning Machine Learning for over six months, and I started to see the term Regressor and can 't really google it out for an exact definetion.

I fully understand logistic regression and linear regression..but just can't figure out this term,..(and sometimes so many terms coming out from ML but they more or less refers to similar things..)

I just got this answer from wiki:

The independent variables, also known in a statistical context as regressors, represent inputs or causes, i.e., potential reasons for variation or, in the experimental setting, the variable controlled by the experimenter.

Is this correct - the learning of ML, to me, really is so much slightly varying terminologies... (which can be completely absent in one course or framework, and suddenly they appear elsewhere..)

r poon
  • 301
  • 1
  • 2
  • 6

2 Answers2

20

Yes, your understanding is correct. Feature, independent variable, explanatory variable, regressor, covariate, or predictor are all names of the variables that are used to predict the target, outcome, dependent variable, regressand, or response. The terminology is ambiguous as it comes from different fields: statistics, econometrics, and machine learning.

Tim
  • 138,066
  • 2
    In first group: add covariate and predictor (both of which are used in various other senses), controlling variable. In second group: add response. Trying to produce a complete list is likely to be futile, but all these terms seem commonly used. – Nick Cox Jun 18 '21 at 08:35
  • @NickCox good point, updated. – Tim Jun 18 '21 at 08:41
  • I might add “factor” to the list of predictor types, though it does imply a categorical variable in particular. – Dave Jun 18 '21 at 09:56
  • @Dave never heard it used & sounds super confusing. As Nick Cox said, there is no point in trying to make the list comperhensive. – Tim Jun 18 '21 at 09:58
  • 1
    Various possible meanings of factor include (1) whatever is produced by factor analysis (long-standing) (2) another name for whatever is not a response (e.g. J.W. Tukey, circa 1977) (3) a specific name for a categorical variable (standard now in much statistical software). – Nick Cox Jun 19 '21 at 12:35
0

An explanatory variable is the variable in its raw form. A regressor is the variable as it appears in a regression equation. Hence, the regressor is the explanatory variable if used as is, but its transformation if any transformations are applied. I think that the term 'regressor' applies specifically to regression equations, and would not apply otherwise.

Raymond Anderson
  • 1
  • 1
    I have often seen the transform of something called an explanatory variable. The gradation between "what others say in various places" and "my terminology which is clear and consistent" is a wide grey zone. – Nick Cox Jun 18 '21 at 08:51
  • Agreed! That explanation was provided by Fox, John [2016] Applied Regression Analysis and Generalised Linear Models, 3rd ed. Sage, Thousand Oaks, CA. It made sense to me, and I adopted that interpretation for a book that I have written. That is, the explanatory variable is that which was delivered, prior to any transformation. – Raymond Anderson Jun 19 '21 at 11:29