How to find a single value which we can consider optimal or most significant from large sampling data?

I have large sampling data generated for 2000 variables with each variable having around 2000 sampling points from a model simulation. For each variable from 2000 sampling points I need to select single significant value. The values range from 1e-7 to 999 but usually these two extreme values have least frequency in data and most of the data falls in the range of 0.00001 - 150. In first attempt, I calculated frequency % of each sampling point and selected values which have frequency of occurrence more than 40%, which worked for about 25% of variables. There are sampling points which fall under <40% frequency range which I didn't use to avoid their statistical significance. In second screening I used standard deviation ≤ 2 which again helped me to get around 5% of the variables. I still have around 70% variables which doesn't fall under these two categories. I don't want to use mean of the the data to get other values. I will highly appreciate if someone can point me the method which I can apply to obtain statistically significant values from the rest of data. I am using MATLAB for simulations and data analysis.