Main outcome using ordered response models.

Notes: Dependent variables are: choice (change in percentage of children choosing the dried apple), liking (change in liking measured on a 4-point scale), and comparison (change in number of times the dried apple is preferred in 5 pairwise comparisons). Independent variables are the binary Treatment-group indicator and baseline values are the respective food valuations in the first assessment; we also control for school specific fixed-effects. Estimates for choice based on a logit model, liking and comparison estimated by an ordered logit. P-values below the coefficients based on clustered standard errors on the class level. P-values below 0.1 in bold. Columns (1), (2) and (3) refer to the second assessment, columns (4), (5) and (6) to the follow-up.

(DOCX)