analyses in longitudinal studies. Prior adiposity is probably the strongest predictor of current or future adiposity. Nonetheless, when early adiposity is taken into account, the positive association between television viewing and adiposity is statistically significant. Because television viewing is significantly associated with adiposity when a variety of other adiposity-related variables, including parental and initial child adiposity, are included in the analyses, the claim that television advertising influences adiposity cannot be rejected on the basis of the current evidence.
The second approach to testing the claim is presented in Table 5-23. Here the number of results for which television advertising was significant and nonsignificant is compared as the number of television-related nonadvertising, third, other adiposity-related, and all variables in the analysis increased. Again, the counts are taken from the statistical analysis that predicted adiposity, included television viewing (the measure of advertising exposure) as a predictor, and had the largest number of television-related nonadvertising, third, and other adiposity-related variables in the analysis. For the television-related nonadvertising variables, as more variables of that type were included, from none to three, the percentage of significant results tended to increase. For third variables, as the number of variables increased from one to four, the percentage of significant results decreased from 100 percent to 57 percent with all four third variables included in the analysis. For other adiposity-related variables, as the number of variables increased from none to two, the percentage of significant results increased somewhat. For the total number of variables, no matter what type, except for the two studies with the largest number of variables (14, 21) in which television viewing was respectively not significantly and significantly associated with adiposity, as the number of variables in the analysis increased from 3 to 12, the percentage of significant results also increased. Overall, then, because in most cases television viewing remained significantly associated with adiposity for the majority of results when more variables were included in the analyses, the claim that television advertising influences adiposity cannot be rejected on the basis of the evidence at hand.
In summary, the 17 results from cross-sectional and longitudinal studies with medium causal inference validity provided useful—but not conclusive—information about the likelihood that the positive association reflected a causal relationship of exposure to television advertising on adiposity. The majority of studies that provided evidence about plausible alternative explanations of the statistically significant positive association between exposure to television advertising and adiposity did not convincingly demonstrate that this association was explained by these alternative variables and not explained by exposure to television advertising. However, it is an unwarranted inference that alternative explanations can therefore be