**Parametric Comparison**

**DTW is an example of one tradition of computational comparison of sounds - estimating distance using contours throughout the element's length. An alternative method is to calculate distance based on a discrete number of scalar parameters, such as "Maximum frequency", or "Length" of the signal. Luscinia makes the measurement of such parameters easy, and provides two ways in which you can analyze sounds based on them. The first is simply to be able to export summary statistics into a spreadsheet or text file, from where you can analyze them using one of any number of statistical approaches. This is dealt with in more detail here.**

The second method is to use Luscinia itself to analyze sounds based on these parameters. Luscinia only provides one multivariate statistical approach to combining parameters - principal components analysis. Dedicated statistics software provides more options. The advantage of carrying out such an analysis within Luscinia is that it can provide you with an interactive visual representation of the results, and built in clustering and syntactical analysis.

Parametric comparison is an option from the second Analysis window page.

The panel that appears when you select this option is basically a large array of different types of parameters for you to choose from. Each row corresponds to a different acoustic parameter (e.g. "Peak frequency"), while each column corresponds to a different statistic associated with that parameter (e.g. "Mean"). So, for example, the "Mean" of the "Peak Frequency" would be calculated as the arithmetic mean of all the peak frequency measures associated with an element. Two of the parameters, “Time” (Length) and "Gap After" only have one statistic associated with them.

After you have selected the parameters that you wish to analyze, you should click "Next Step". At this point, Luscinia calculates a distance matrix based on these parameters for every pair of elements in the comparison scheme you have selected. Each parameter is normalized by its standard deviation calculated over all elements in the comparison. The distance between two elements is then calculated as the Euclidean distance over all of the selected parameters.