Carbon nanotube (CNT) diameter control is essential for many conductive and structural applications; to date, a general diameter methodology applicable to all floating catalyst chemical vapor deposition (FC-CVD) reactors is not yet established. Single-walled CNT (SWCNT) diameter was robustly controlled by systematically exploring an extensive variety of fundamental FC-CVD reactor parameters (631 experiments) including: CO2, H2, and Ar flowrates; fuel flowrate; added H2O (up to an unprecedented 30% of fuel by mass); ferrocene concentration; reactor temperature (covering temperatures with solid and liquid floating catalyst); and quartz reactor tube age. As a function of these experimental input factors, across three Raman wavelengths, validated statistical models robustly predicted: 1) diameter of SWCNTs, via the weighted average of the radial breathing modes (RBMs); 2) the conditions which led to SWCNT growth, via the categorical appearance of viable RBMs. This is noteworthy because it illustrates a new paradigm of FC-CVD reactor control, accounting for interaction between reactor input factors, system noise, and other non-linear responses. Oxidative influences (including laboratory humidity) yield smaller CNT diameters, although H2O becomes important only at higher temperatures and CO2 is countered by H2 addition. Increasing reactor temperature increased diameters. Pyrolysis calculations showed species inside the reactor differ substantially from species injected. A kinetic-thermodynamic model of catalyst particle size and chemical state revealed that adding oxidative species kept the active catalyst particles small, explaining the resulting small CNT diameters. The powerful combination of our multivariate statistical approach with kinetic-thermodynamic modeling enables unprecedented insight and control over CNT synthesis.