User blog comment:Sirebel/Mathematician Required - Apply Within/@comment-27170954-20170427155020

Sirebel, et-al,

I ran a regression on the data you provided. Here is what it says:

Base: 0.0465 ± 0.0529

Engine: 0.2897 ± 0.0376

Drivetrain: 0.2692 ± 0.0404

Body: 0.2408 ± 0.0361

Suspension: 0.3413 ± 0.0617

Exhaust: 0.3186 ± 0.0625

<p class="MsoNormal" style="margin-bottom:0in;margin-bottom:.0001pt;line-height: normal"><span style="font-size:12.0pt;font-family:"Times New Roman",serif; mso-fareast-font-family:"Times New Roman"">Brakes: 0.1939 ± 0.0358

<p class="MsoNormal" style="margin-bottom:0in;margin-bottom:.0001pt;line-height: normal"><span style="font-size:12.0pt;font-family:"Times New Roman",serif; mso-fareast-font-family:"Times New Roman"">Tires & Wheels: 0.2698 ± 0.0506

<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto; line-height:normal"><span style="font-size:12.0pt;font-family:"Times New Roman",serif; mso-fareast-font-family:"Times New Roman"">These are 95% confidence intervals and this assumes all levels within a category are the same, which may or may not be true. The width of the interval is large – greater than 0.1 in some cases. This is because the dataset is small relative to the number of variables. Also because the dataset is small, there are still errors in using these values for prediction – 5 of the 28 estimates (when rounded to 1 decimal) still differ from the actual value for PR by 0.1. Truncating produces 13 predictions that are off by 0.1.

<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto; line-height:normal"><span style="font-size:12.0pt;font-family:"Times New Roman",serif; mso-fareast-font-family:"Times New Roman"">To improve the estimate, more data is needed. I built a spreadsheet with 12 specific upgrade paths (one of which you have already completed). The paths chosen attempt to get a minimum of 260 data points while balancing the participation of each of the 26 individual upgrades. Because of the unavoidable duplication of some data points, 12 paths produces 269 unique samples (not 12x26=312). I’m not sure I achieved the second goal, but I do think I’ve included each upgrade frequently enough to make for decent estimates.

<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto; line-height:normal"><span style="font-size:12.0pt;font-family:"Times New Roman",serif; mso-fareast-font-family:"Times New Roman"">With this data, I hope to be able to state if each level in a category is the same PR increase or not. I should also be able to test the ability to predict an upgraded PR more accurately.

<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto; line-height:normal"><span style="font-size:12.0pt;font-family:"Times New Roman",serif; mso-fareast-font-family:"Times New Roman"">What I need help with is getting the data. You used the new NISSAN Skyline 2000 GT-R (KPGC10), which is still locked for me and I have no idea how to unlock it to help with the data collection (I also only have one device, so am reluctant to manipulate it too much for fear of permanently impacting my game). Can you and some of your admin brethren with the ability to preview the car tackle the specified upgrade trees? Then I can run the regressions and report back what I find?

<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto; line-height:normal"><span style="font-size:12.0pt;font-family:"Times New Roman",serif; mso-fareast-font-family:"Times New Roman"">Here is the data file: https://drive.google.com/file/d/0B0W9o_brxvhEQWxfcjZ6dlNWaDA/view?usp=sharing