This report analyzes data gathered from an experiment that was carried out in the 2020 Census. A new training module was designed for bilingual enumerators to aid in administering the census questionnaire to Spanish-speaking households in the nonresponse follow-up operation. An objective is to study the association between the training and response rate. A multinomial regression with continuation-ratio logit link is used to model the sequence of contact attempts to affected households. Several log-odds ratios are considered to provide inference about this relationship. We do not find significant evidence to conclude that the training is associated negatively with response rate; i.e., training appears not to harm response rate. We suspect that a stronger conclusion can be made about the training and improvement to response rate in applicable contact situations; however, a more controlled experiment with associated data would be needed to make this assessment. Additionally, a power analysis which was carried out in advance of the experiment is now revisited. Here, power of the selected log-odds ratios is evaluated under the experiment as carried out in practice using the realized sample size.