The values specified for number of poles and zeros must match the input-output dimensions of estimation data. You can specify them as scalars if the value is the same for all input-output pairs.