Calculate slope in model for a 12 unit change in x

[Jon Heintz]

I have the following model

glm LOS c.time##i.group, f(gam) l(log)

where

• LOS is continuous, skewed right
• time is continuous in months.
• group is categorical taking on values 1, 2, 3.

I found the average slope of LOS/month for each center, so I have

margins, dydx(time) at(group=(1,2,3))

After running this, I found the slopes are too small to be interpretable, so I would like to change the units of time to be years rather than months. Is there a way to do this without creating a new time variable and rerunning the model?

[Joseph Coveney]

Maybe

display in smcl as text "Slope in years = "  r(table)["b", "time"] * 12 " ± " r(table)["se", "time"] * 12

[Joseph Coveney]

Note in edit: you'll need to do that separately for each level of the group variable. Something like

forvalues group = 1/3 {
    display in smcl as text "Slope in years = "  r(table)["b", "time:`group'._at"] * 12 " ± " r(table)["se", "time:`group'._at"] * 12
}

[Dimitriy V. Masterov]

You can use a contrast of margins. Example here.