regression with "grouped" variables.

[Marcel Meyer]

Hello,
First I'd like to apologize for the probably rather beginner question I'm going to ask, but I'm really new to stata.
Second:
I have a problem concerning a regression I'm trying to do:

I have datas about subjective well-being, Income, age and so on.
To get a better view about the correlations between these variables, I tried to, for example, group "age" in groups from 25-30,31-35, 36-40
by using commands as:

. generate low_age1 = Age if Age >= 25 & Age < 30.

Now my problem is that when I'm trying to make a linear regression as following:
. regress SWB low_age1 low_age2
for example, it just tells me that there are "no observations r(2000)";

If I make other groups including the previous ages with commands as:
. generate low_age1 = Age if Age < 30
. generate low_age2 = Age if Age < 35

It gives me a regression but says "low_age2 omitted because of collinearity" and therefore doesn't show me a coefficient for "low_age2".
I hope this kind of makes sense and is understandable.
I'd be really greatful, if someone could help me with this probably newbish question.
Thanks a lot in advance
Marcel M.

[Carlo Lazzaro]

Marcel:
welcome to the list.
The issue there seems to be related to (too many) dummies created mistakenly.
You should be better off with -fvvarlist- and -label- for categorizing age (please note that you should categorize -Age- for all the values included in your dataset, otherwise Stata will listwise delete those observation with missing -low_age-) as :

generate low_age = 0 if Age >= 25 & Age < 30
replace low_age=1 if Age >= 30 & Age < 35
replace low_age=2 if Age >= 35 & Age <= 40
label define low_age 0 "25-29" 1 "30-34" 2 "35-40"
label val low_age low_age
regess SWB i.low_age <other_controls>

[Marcel Meyer]

Marcel:
welcome to the list.
The issue there seems to be related to (too many) dummies created mistakenly.
You should be better off with -fvvarlist- and -label- for categorizing age (please note that you should categorize -Age- for all the values included in your dataset, otherwise Stata will listwise delete those observation with missing -low_age-) as :

generate low_age = 0 if Age >= 25 & Age < 30
replace low_age=1 if Age >= 30 & Age < 35
replace low_age=2 if Age >= 35 & Age <= 40
label define low_age 0 "25-29" 1 "30-34" 2 "35-40"
label val low_age low_age
regess SWB i.low_age <other_controls>

This has solved my problem really fast, thank you very much.
I wish you a happy easter !
Kind regards,
Marcel

[Carlo Lazzaro]

Marcel:
you're welcome.
I do reciprocate the same to you and your dears.