Help with rescaling

[Al Bothwell]

Hello Everyone,

I am fairly new to Stata so this question might be basic. In a dataset, I have age in continuous form (for example, 25, 26, 28, 32, 36 and so on). I would like to re-scale age into 5 year increments and then run a linear regression using the re-scaled variable. How do I implement the re-scaling in Stata. Your help would be much appreciated.

Al Bothwell

[Nick Cox]

Why throw away information? Let's suppose I am 16. You could round my age to 15 or to 20. Either way, I object, and more importantly, why throw away information?

[Maarten Buis]

I agree with Nick if the intend was to round the age. An alternative I use often is to just change the unit by dividing the old age variable by 5. So in a regression analysis we see the effect of getting 5 years older.

[Phil Bromiley]

While agreeing with Nick and Maarten, to directly answer your question, you could use recode to do the recoding. There are more elegant ways to do it, but recode works.

[Nick Cox]

I dislike recode because, recursively, I dislike recode and don't usually use it, or exceptionally have to relearn some of its syntax on the very odd occasions I try to debug someone else's recode code. That is purely personal and not worth a microhill of microbeans.

But I think it's worth pointing out ways of rounding in 5 unit intervals:

 
generate newto = round(old, 5) 
generate newup = 5 * ceil(old/5) 
generate newdn = 5 * floor(old/5)

I've asked in the past for

 
* syntax not supported yet! 
generate newup = ceil(old, 5) 
generate newdn = floor(old, 5)

One enormous advantage of any of these is that the new values are self-describing, given the function used.

[Al Bothwell]

I ended up using what Maarten suggested:

generate age5=age/5

Thus, a unit change corresponds to five years. This is exactly what I wanted.

Thank you everyone!
Al Bothwell