the estimated overall regression relationship £(my) = a0 + the estimated child-specific regression relationship in the kth child E(mkj) = a0 + uk + {fio + vk)skj; and the modelled 2-5 per cent and 97-5 per cent percentiles for the observed meter readings in the fcth child calculated as

E(mkj) ± 1-96 x J{a2e + 2&efskJ + a}s2kj). (8) Interpreting the model. Having used the multi-level model to derive values for the statistics of principal interest there are a number of approaches which one might take to the combination of information about the magnitude of random within-child variation and the magnitude of the m:s regression gradient in order to draw conclusions about the potential utility of portable PEF monitoring. We will restrict our attention to one possible approach. Although we believe this method to be sensible, we do not claim that it is 'optimal' in any technical sense of the word.

In order for a child to respond appropriately to a reduction in meter PEF following a period of stable true (spirometric) PEF, it is necessary for the observed change to be discriminated from the background noise representing random within-child variation of m around its expected value during the period of stability. On the basis of an empirical review of a series of computer e

Was this article helpful?

0 0

Post a comment