For fluorescent lamps, a negative binomial regression was conducted.4
Results are reported in Table 5. The chi-square test shows
that the model is statistically significant at the 99 percent level. As
above, since each of the independent variables is a categorical
variable, the coefficients should be interpreted as the change in
number of lamps collected compared to the omitted category.
Again, the results show that container convenience and the type of
store are of the greatest statistical significance. Stores where the
collection container was least convenient were likely to collect 3
fewer lamps than a similar store with a prominently placed
collection container. Hardware stores collected the most lamps,
with the coefficient on all but the mass merchant stores being
negative and statistically significant. The coefficients on the variables
indicating the status of the stores were negative but not
statistically significant. None of the coefficients on the parking
variables or the hours variable were statistically significant