anti-tobacco messages.
This spreadsheet enabled us to easily view the data segments and the codes applied to them in analytically useful ways. First, the transcripts were read in their original, sequential context. The segments were indexed into topical codes enabling us to see horizontal patterns in the co-occurrence of topics for each data segment. An initial 53 topical codes were identified from the transcripts and through discussions among the three coders the code list was subsequently reduced iteratively to 42 as we combined topical codes with significant overlap (Table 1). By using their initials when coding segments, coders were able to identify and discuss discrepancies in how each coder applied codes to each data segment until we reached agreement.