I have been working with guidelines for reference intervals suggested by the ASVCP (https://www.asvcp.org/pubs/pdf/RI%20Guidelines%20For%20ASVCP%20website.pdf).  They recommend non-parametric methods when samples size is >120.  If transforming you data is  not an option then a bootstrap method may work to determine you RI.  If you or someone you know uses R for stats I recommend the "referenceintervals" package (https://cran.r-project.org/web/packages/referenceIntervals/referenceIntervals.pdf) which offers all of these options.

Thank you, Nichols, for your valuable suggestions and information. Yes, ASVCP recommends non-parametric methods but it requires sample size no less than 120 whereas, mine is between 90-100. It looks like robust technique would be preferable (especially in non-gaussian distributions) and for other cases, non-parametric bootstrap will serve the purpose. I'm using 'reference value advisor' MS excel plugin and it has all these options.