Regression on different sub-groups

[Anh Quynh Nguyen]

Hi all,

This is the dataex of my dataset. I am working on the impact of drinking alcohol on earnings with the following variables:

input double days_drink_wk float drink_intensity byte educgrp long earnings
.5 4 1 5000
.5 .5 3 20000
.5 .5 3 24000
1 1 4 101099
2 12 3 11000
3 30 3 30000
3 6 4 100000
.5 .5 2 24000

• days_drinks_wk: numbers of days drinking per week
• drink_intensity: numbers of drinking units per week
• educgrp: education level
• earnings: annual income in $$$

So I would like to categorise them into different subgroups in terms of drink_intensity (0: abstainers, 1-7: light drinkers, 8-15: Moderate drinkers,...) and run a regression to find how these different groups affect earnings. I also want to include demographic variables like age, education, marital status. Please find the example below:

The way I do it is generate new variables for each of the group and run individual regression, as below:

generate abstainers = drink_intensity if inrange(drink_intensity, 0, 0.5)
generate light_dr = drink_intensity if inrange(drink_intensity, 0.6, 7)
generate light_moderate_dr = drink_intensity if inrange(drink_intensity, 7.1, 21)
generate moderate_dr = drink_intensity if inrange(drink_intensity, 21.1, 43)
generate moderate_heavy_dr = drink_intensity if inrange(drink_intensity, 43.1, 64)
generate heavy_dr = drink_intensity if inrange(drink_intensity, 64.1, 86)
generate very_heavy_dr = drink_intensity if (drink_intensity > 86)

reg earnings abstainers educgrp married

All of the code work, but I don't think it is the correct way to do because it will generate different coefficients of the demographic variables for each group. So my questions are:

1. Do I run it correctly, and if not, could you let me know the correct code? Just one two example and I could work out the other group.
2. Should I take log in the earnings, currently earnings are in $ but I want to interpret data as % of earnings increase as we drink more or less. If yes, could you let me know the code?

Thank you very much!

[Carlo Lazzaro]

Anh:
1) I would create a categorical variable including all the subgroups of drinkers and plug it in the right-hand side of the regression equation (see -fvvarlist- and -label-);
2)

gen ln_earnings=ln(earnings)

[Anh Quynh Nguyen]

Anh:
1) I would create a categorical variable including all the subgroups of drinkers and plug it in the right-hand side of the regression equation (see -fvvarlist- and -label-);
2)

gen ln_earnings=ln(earnings)

Thank you so much. It works. I am wondering if I want to do separate regression for male and female (as in the screenshot above), should I run the gender variable as a factor variable, or should I use the if command? Many thanks!!

regression ln_earnings i.drink_intensity i.gender

or

regression ln_earnings i.drink_intensity if gender == 1

[Carlo Lazzaro]

Anh:
I would add -i.gender- in the right-hand side of my regression equation.
Then I would -predict fitted, xb- the fitted values for male and female.

[Anh Quynh Nguyen]

Thank you so much Carlo. Got it!!!