The model object can even store the data used for identification, in the property data. Therefore, plot(idmobj) can plot not only the transfer function, but also the measured data, and the error of identification.
load bandpmod get(bkfit) version = 1.2 date = '22-Oct-1998 09:09:26' history = {[1x48 char]} data = [1x1 fiddata] variable = 's' num = [-4.0066e-18 -5.1034e-13 -9.2494e-11 2.7921e-09 1.5859e-05] denom = [-9.4278e-23 -2.3515e-19 -3.3573e-15 -5.1609e-12 -3.3479e-08 -2.3288e-05 -0.093552] representation = 'polynomial' freqvect = [16x1 double] fscale = 3835 delays = 0 covariance = [13x13 double] fitinfo = [18x1 double] >> set(bkfit) name: string version: version number, set by the system date: string (date + time) notes: string history: cell vector of strings data: optional, input data of estimation (tiddata or fiddata) algorithm: structure describing the algorithm userdata: user-defined variable: [ 'z^-1' | 's' | 'r' | 'w' ] representation: [ 'polynomial' | 'orthopol' ] num: cell array of numerators denom: cell array of denominators ntr: cell array of transient numerator polynomials Znum: array of weight vectors of numerator Zdenom: array of weight vectors of denominator Zntr: array of weight vectors of transient numerator freqvect: column vector or array, freqs x channels chnames: string cell array, length: channels chtypes: string cell array, [ 'input' | 'output' ] fscale: scalar, optimum scaling frequency delays: vector of delay values units: cell array of strings covariance: array of covariances fixedpar: nx2 array of fixed parameters fitinfo: cell array of information on fit