What does the jargon "constant regressor" mean in Stata?

[Tiago Pereira]

Dear Statalisters,

"Regress body mass index (bmi) in kg/m² on age and a constant regressor."

In Stata, do you know what does the "constant regressor" mean? Or is the reviewer mistaken in that term?

My dataset looks like:

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(age bmi)
86   27
66 26.2
51 24.8
11 24.2
30   25
62 25.9
40   26
55   25
18   23
25 24.6
67 24.8
72 28.5
38 24.3
60 27.3
26 24.9
60 25.6
36 24.1
80 25.4
17 24.1
33 25.9
end

Any thoughts? Comments?

All the best,

Tiago

[Jared Greathouse]

The Beta_0 term/constant/y-intercept. They're likely just being super technical (as far as I know), since normal people would just say "Estimate the impact of x on y using OLS with a constant term." Or other words to that effect, assuming I understand them well

[Leonardo Guizzetti]

I agree with Jared that this only makes sense if the reviewer is referring to the intercept, which would have been included by default unless you specifically suppress it from estimation using the nocons option. There’s a question of whether the constant aids interpretation if model really is this simple (I suspect it isn’t, and gender is missing from the model at minimum), no tensoriale have been applied, and the model is presented only in terms of the estimated coefficients. For example

regress bmi age

the model intercept tells you the grand mean BMI when age is zero, which is clearly absurd. However, if you were to present graphical aids along with the model to show how BMI varies with age, this helps. But if you centre age at some reasonable value, then direct interpretation of the model intercept is sensible.

[Tiago Pereira]

Thank you, Jared and Leonardo, for the tips and comments. Very useful. So, basically what the reviewer is asking is:

regress bmi age, nocons

?

All the best,

Tiago

[Leonardo Guizzetti]

Thank you, Jared and Leonardo, for the tips and comments. Very useful. So, basically what the reviewer is asking is:

regress bmi age, nocons

?

All the best,

Tiago

They’re asking for the exact opposite, they want the intercept in the model so take out the nocons option.

[Carlo Lazzaro]

Tiago:
in 1998, the deeply missed Doug Altman and Martin Bland published a paper on this topic on BMJ (Generalization and extrapolation is its title; I miss the full reference at the moment).
The article can be downloaded free of charge.

[Carlo Lazzaro]

Tiago:
the link to the article is https://pubmed.ncbi.nlm.nih.gov/9694763/.

[Tiago Pereira]

Got it!

As always, thank you much, folks, for the help and extremely helpful tips and references. Thank you very much Jared, Leonard and Carlo!

All the best,

Tiago