Well, I'll stay on the post topic. It's unlikely that you can get the lab notebooks from most of 20th century Physics, this is what you're really looking for. It's not that important, because once the results are published by a few different groups that don't trust each other and will scrutinize their peers' work, all relevant data have been measured again and again. If theories like Special Relativity were really wrong there would be consequences like having to close all particle accelerators because they wouldn't work. Many key experiments are routinely done as part of the university courses, and many results are also routinely used in other fields, engineering included. E.g. linearisations of Planck's distribution are done in Remote Sensing for estimating surface temperatures, they work great btw. We also understand light bulbs, pyrometers, infrared thermometers... -those work straight with Planck's Law, how else could they find the temperature-. There are GR corrections coded somewhere for your GPS to work, etc.
If by seeing "how well different theoretical models express experimental data" you're planning to apply idk assorted non-linear regression models from equations you're not familiar with to see which one has the minimum chi-square or something like that, well I guess that's what hobbies are for. Differences tend to appear at the limiting cases, there you need more Physics than Levenberg–Marquardt fun, so it's better to start there.
I have to warn you that your wording matches the usual stuff you get from Physics cranks (the bane of the trade), hence the knee-jerk reaction, sorry about that. I'll try harder next time.