How to find the proportion of participants that improve by a certain amount

[Erica Ramsden]

I need to find the proportion of participants within a group that have improved by a certain amount. Within my dataset, I have the variables legpainwalk0 (baseline) and legpainwalk3 (52 weeks) with scores for these groups ranging from 0-10. I want to find the proportion of people that have improved by 2 points or more from baseline however, I am unsure of how to do this, any help would be greatly appreciated.

Thanks! Erica

[Nick Cox]

Sounds like

bysort group : egen wanted = mean(cond(!missing(legpainwalk0, legpainwalk3), (legpainwalk3 - legpainwalk0) >= 2, .))

See also https://journals.sagepub.com/doi/pdf...867X1101100210 especially Section 9.

The logic is that (legpainwalk3 - legpainwalk0) >= 2 evaluates as true (1) or false (0) and that the mean of a (0, 1) indicator is a proportion. We just need to make sure that missing values are treated separately.

[Erica Ramsden]

That's worked, thank you so much for the help!

[Erica Ramsden]

Sounds like

bysort group : egen wanted = mean(cond(!missing(legpainwalk0, legpainwalk3), (legpainwalk3 - legpainwalk0) >= 2, .))

See also https://journals.sagepub.com/doi/pdf...867X1101100210 especially Section 9.

The logic is that (legpainwalk3 - legpainwalk0) >= 2 evaluates as true (1) or false (0) and that the mean of a (0, 1) indicator is a proportion. We just need to make sure that missing values are treated separately.

I also have to do this for the variables maxwalkingdistancem0 and maxwalkingdistancem3 but see the proportion of people who had an improvement of 20% from baseline how would you recommend to find this?

[Nick Cox]

How is that measured? What happens if the baseline is zero?

[Erica Ramsden]

This was a self-report measure with participants reporting in metres how far they are able to walk. The lowest score at baseline was 2m

[Nick Cox]

So you want the mean of

cond(!missing(maxwalkingdistancem0, maxwalkingdistancem3), maxwalkingdistancem3 >= 1.2 * maxwalkingdistancem0), .)

presuming that 20% improvement means 20% improvement or more.