The registration of functional brain data to the common brain space offers great advantages for inter-modal data integration and sharing. However, this is difficult to achieve in functional near-infrared spectroscopy (fNIRS) because fNIRS data are primary obtained from the head surface and lack structural information of the measured brain. Therefore, in our previous articles, we presented a method for probabilistic registration of fNIRS data to the standard Montreal Neurological Institute (MNI) template through international 10-20 system without using the subject's magnetic resonance image (MRI). In the current study, we demonstrate our method with a new statistical model to facilitate group studies and provide information on different components of variability. We adopt an analysis similar to the single-factor one-way classification analysis of variance based on random effects model to examine the variability involved in our improvised method of probabilistic registration of fNIRS data. We tested this method by registering head surface data of twelve subjects to seventeen reference MRI data sets and found that the standard deviation in probabilistic registration thus performed for given head surface points is approximately within the range of 4.7 to 7.0 mm. This means that, if the spatial registration error is within an acceptable tolerance limit, it is possible to perform multi-subject fNIRS analysis to make inference at the population level and to provide information on positional variability in the population, even when subjects' MRIs are not available. In essence, the current method enables the multi-subject fNIRS data to be presented in the MNI space with clear description of associated positional variability. Such data presentation on a common platform, will not only strengthen the validity of the population analysis of fNIRS studies, but will also facilitate both intra- and inter-modal data sharing among the neuroimaging community.