To estimate the collimator scatter factor, S_c of small field sizes in which a mini-phantom cannot be fully included at the nominal treatment distance (NTD=100 cm), we measured the in-air output of 4 MV and 10 MV X-rays of a Varian’s Clinac 2100 C/D using a mini-phantom at NTD and at a long source-to-chamber distance (SCD=200 cm) with field-size defined at the isocenter down to 4.6×4.6cm² and 2.3×2.3cm², respectively. We then compared the fitted curve to the NTD dataset by a cumulative exponential distribution model with that by a cumulative Gaussian distribution (error function) model containing a zero-field extrapolated term derived from the long SCD dataset. The results showed that the zero-field extensions of two fitted curves coincided for a 4 MV X-ray, but a large discrepancy was seen between them for a 10 MV X-ray. Therefore, the S_c of small field sizes not measurable using a mini-phantom at the NTD can be well estimated by applying the cumulative exponential model to the NTD dataset in the case of a 4 MV X-ray beam filtrated with a cone-shaped flattener. However, to estimate the S_c of such small field sizes in the case of a 10 MV X-ray beam filtrated with a bell-shaped flattener, we consider it preferable to also measure in-air output at a long SCD and to apply the cumulative Gaussian model as described here.