3 The method of the analysis of multicolour photometric data

For the analysis of colour behaviour we use the method first proposed by Choloniewski (1981) and considered recently in detail by Hagen-Thorn (1997). This method has been used by us several times, in particular, to study the previous outbursts of OJ 287 (Hagen-Thorn 1987; Hagen-Thorn et al. 1991,1994).

In the frame of a two-component model (constant component + variable source) the method allows to obtain the relative spectral energy distribution of the source responsible for variability if the source has variable flux but unchanged spectral shape. If this is the case, the points corresponding to the observed fluxes at different times should lie on a straight line in the n-dimensional flux space {F₁,... F_n} (where n is the number of photometric bands) and consequently on straight lines in the planes {F_i,F_j}. Their slopes are the flux ratios of the variable component in the bands under consideration. Plotting these ratios as a function of frequency gives the spectral energy distribution of the variable component in relative units.

Thus if observational points in fact lie on the straight lines within observational errors one can conclude that the model of single variable source with unchanged spectral energy distribution is valid. The slopes of the lines (fitted by the method of orthogonal regression (Wald 1940) used in the case of accidental errors existence in both axes) give above mentioned ratios (the relative spectral energy distribution).