I have a time series which is discreet and bounded (between 0 and 100). It’s actually also increasing over time, from low to high. When I test for a unit root, I find it.
And yet I wonder if unit root tests on bounded variables even make sense?
And if unit-root testing doesn’t make sense in that case, what’s left of the whole cointegration framework? For instance, If I wanted to know if variable Y (0-100) is affected by another bounded variable X (say between 0-10), is it even possible to speak meaningfully about this?
