Ceil function for decimals?

[Sean Tanner]

I want to store p-values in macros to display in graph captions and am having trouble with p-values that are nearly zero as the round(p-value, .001) function returns 0, which looks incorrect in the output:

mgof variablex, mc ksmirnov

global ksfull=round(r(p_ksmirnov), .001)

twoway...caption("KS test rejects null, p=$ksfull")

In many cases I get p=0, where I would prefer to have p<.001. If the ceil() function had the option of specifying digits, as round() does, then this would be trivial. Any suggestions?

Sean

[Nick Cox]

You have not explained mgof as a user-written program. Please see the FAQ Advice.

The function round() produces another number, not a string display. Note that round() with second argument 0.001 can at best produce binary approximations to multiples of 0.001.

You may get closer to what you want with

global ksfull = cond(r(p_ksmirnov) < 0.001, "< .001", string(r(p_ksmirnov), "%4.3f"))

[Sergiy Radyakin]

Sean, perhaps you could apply scaling/subscaling. Compare:

local precision=0.001

display round(0.0002, `precision')
display round(0.0008, `precision')

display ceil(0.0002/`precision')*`precision'
display ceil(0.0008/`precision')*`precision'

Produces values 0 and 0.001 for round() function and 0.001 and 0.001 for the ceil() function.

Best, Sergiy Radyakin