Dear forum members,
I would be very thankful if you could help me ... is there a step-by-step guide to tune a model?
I have three blocks of variables (situation, strategy, result). Situation has 6 formative LVs, strategy 5 formative LVs and result 3 reflective LVs. I want to analyse the causality of the situation over the strategy, the situation over the result and the strategy over the result. At the end, I would like to formulate statements like “In a Situation A, using Strategy B, we obtain the result C”. Furthermore, I want to compare two data sets (Mexico and Germany)
I imported my all data set for Mexico. I created the whole model. That is, each situation is linked to every strategy and to every result, and each strategy is linked to every result... then, I ran the PLS procedure and I got some values. I got some negative outer loadings for my Lvs, some low path coefficients between LVs. Therefore, I deleted the MV with negative outer loadings. I ran once again the PLS procedure and I deleted the path coefficients which had an absolute value lower than 0,1. I ran once again the PlS procedure ... and I got new (negative) values in the outer loadings of the MVs. Which troubles me ... because if the outer loadings vary (that much) every time I make a change, then I will never be able to tune my model.
As alternative I create a model with all the situations and strategies but only with one result (I wanted to analyse the causality of the situation/strategy separately for each result). But then I got (very) different outer loadings for the situation/strategy LV in each model. Which means that It wouldn't be possible to talk about the “same” situation/strategy LV...
I am sure I am missing something... is there a step-by-step guide to tune a model?
I really appreciate your help.
Best regards,
Atl