Hey guys,
this question is rather practical: I'm using SmartPLS to verify existing psychological scales and their impact on marketing related constructs. Usually, if an indicator loads poorly onto a construct (loading below .4) we should delete it, right? Right! But how do I go about interpreting the results if they were obtained by means of already well established scales? Literature says, those scales work well. however, single indicators do not work well in my PLS-SEM model with my data. Should I delete the indicators to get satisfactory values for AVE, path coefficients, and so on or should I just disregard the whole thing, saying "the original scales do not work properly with the data", thus no analysis possible.
How would you think about this issue if you had a paper to grade or gave feedback to an author for publication purposes?
Thank you so much for advice!