一段话,看不懂
谁来帮忙解释一下
6.3 The estimation-detection dilemma
A crucial point when analysing functional imaging data (PET and fMRI alike) is that in general the model is not well known. The larger the model, the better in general would be the estimation of the signal, as long as the model is not starting to capture noise components. This often leads to less specific questions and to less sensitive tests compared to situations where the difference is known with better precision. There are two extreme choices:
> the use of a simple model with the danger of not having modelled some effects properly, a situation that may lead to biased results
> the use of a very flexible model with less sensitive tests and difficulties in the interpretation of the results
In other words, it is difficult to estimate the signal and at the same time test for this signal. A possible strategy would consist in using part of the data in an estimation phase that is separate from a testing phase. This will, however, involve ”losing” some data. For an instance of such a strategy see [3].
6.3 The estimation-detection dilemma
A crucial point when analysing functional imaging data (PET and fMRI alike) is that in general the model is not well known. The larger the model, the better in general would be the estimation of the signal, as long as the model is not starting to capture noise components. This often leads to less specific questions and to less sensitive tests compared to situations where the difference is known with better precision. There are two extreme choices:
> the use of a simple model with the danger of not having modelled some effects properly, a situation that may lead to biased results
> the use of a very flexible model with less sensitive tests and difficulties in the interpretation of the results
In other words, it is difficult to estimate the signal and at the same time test for this signal. A possible strategy would consist in using part of the data in an estimation phase that is separate from a testing phase. This will, however, involve ”losing” some data. For an instance of such a strategy see [3].
评论
发表评论