The Monte-Carlo experiment with the least squares methods of line fitting

Different techniques of fitting the straight line to TL data points are compared in this paper. In the experiment "artificial" data points are generated by means of the Monte-Carlo method according to three models: (i) the dispersions of TL values are the same for all points; (ii) the dispersions are the fixed percentage of the expected TL values; (iii) the mixture of models (i) and (ii). Straight lines are fitted to each set of data points by means of different regression techniques: TL on D, D on TL, the main axis, unweighted and weighted in different ways. The comparison of results, i.e. values of ED and δED, obtained by means of these different techniques enables the conclusion that the most suitable technique is TL on D regression weighted with sample variances.