I don't know your specific case but, typically, when reviewers ask for analytic evaluations they want numerical data and a repeatable procedure used to find them. These data allows you and the reviewers to have numerical data related to your results and to compare your approach with other similar recent approaches showing advantages and disadvantages of each approach. 

In a first step t is required to define some measures which can strengthen your research; then it is needed to calculate them on one or more specific case studies and compares results with other approaches. In order to have meaningful data it is needed to repeat the experiments many times, then the result can be showed as mean, median and standard deviation of each measure.

Below three papers including these data.

In other cases, when the research led to perception results, a survey can be submitted to a group of people that evaluate the developed system and answers to a questionnaire. I have seen this approach in many Virtual Reality papers on visualization perception and I have adopted a similar approach in a recent paper (see last link).

Article The Improvement of Test Case Selection for the Process Softw...

Article Comparing the Effectiveness of Software Testing Strategies

Article Variation factors in the design and analysis of replicated c...

Conference Paper A Virtual Prototyping Platform to Improve CAE Analysis Workflow

Hi, you can see the book "Empirical Software Engineering and Verification" of Bertrand Meyer, maybe you obtain some ideas. 

Typically in Empirical analysis, first of all you have to carry out Validity and Reliability Analysis of your data. Correlation of the  data need to be computed ( Pearson and Spearmen Correlation techniques could be used depending upon whether the data is random or ranked). Depending upon the size of the test data, internal and external validity tests need to be carried out as well.

One option is to collect some empirical data of the proposed testing technique, for example, by means of a controlled experiment where some units be exposed to the proposed technique and some units act as control. In this way you can assess a possible effect of your technique. After that, you can contrast your evidence with other works.

Hi, I do not know exactly your case but I think that some paper published in "JOURNAL OF TESTING AND EVALUATION" can give you answer. 

Your obtained results of testing technique should be modeled as a semi-empirical proposing the assumptions. The assumptions if you can quantify by a parameter in form of equation gives you a beautiful 'Analytical expression'. Priory, concern be given for the validation of your test data before forming it as a model as discussed above.