@@mronkko What about the CR? if 1 out of 3 indicator is positive and the other 2 is negative then the sum of factor loadings will be negative or very small, should I make it into absolute value for the 2 negative loadings?
@@fianathalie4837 That deepens. If some of the items are reverse scored, (e.g. I feel good, 1=disagree, 5= agree and I feel bad, 1= disagree, 5 = agree) then reversing the indicator coding prior to analysis would be your best choice (in my opinion) and would make all the loadings to be of the same sign. If all the items are coded the same way (i.e. none are reversed) but some loadings are negative, that indicates a model specification problem and you probably should not trust any reliability statistics calculated from those data.
Professor, 1.- I thought that CR was calculated only with CFA. 2.- If a construct has three dimensions and I check it with the EFA, does that indicate that each dimension is one-dimensional? and should apply a CR for each dimension? Thank you very much!
1) The difference between EFA and CFA is smaller than what most people think. 2) Unidimensionality must be established for each dimension and you need to estimate reliability for each dimension separately, unless you want to go for multidimensional reliability coefficients. Cho, E. (2016). Making reliability reliable: A systematic approach to reliability coefficients. Organizational Research Methods, 19(4), 651-682. doi.org/10.1177/1094428116656239
This is the guy who makes 10 professors who are PLS-SEM specialists to respond to his paper while he is a student doctor in the year 2013.
I love your channel! Really, your videos are so well researched and explained! Thank you so much!
You are welcome!
Good explanation. Thank you!
If one of my indicator has negative loading factor, should I make it to absolute value to calculate the AVE?
Hi. AVE is weighted sum of squared factor loadings so the signs do not make a difference. But in practice we would not take absolute values.
@@mronkko What about the CR? if 1 out of 3 indicator is positive and the other 2 is negative then the sum of factor loadings will be negative or very small, should I make it into absolute value for the 2 negative loadings?
@@fianathalie4837 That deepens. If some of the items are reverse scored, (e.g. I feel good, 1=disagree, 5= agree and I feel bad, 1= disagree, 5 = agree) then reversing the indicator coding prior to analysis would be your best choice (in my opinion) and would make all the loadings to be of the same sign. If all the items are coded the same way (i.e. none are reversed) but some loadings are negative, that indicates a model specification problem and you probably should not trust any reliability statistics calculated from those data.
Professor,
1.- I thought that CR was calculated only with CFA.
2.- If a construct has three dimensions and I check it with the EFA, does that indicate that each dimension is one-dimensional? and should apply a CR for each dimension?
Thank you very much!
1) The difference between EFA and CFA is smaller than what most people think.
2) Unidimensionality must be established for each dimension and you need to estimate reliability for each dimension separately, unless you want to go for multidimensional reliability coefficients.
Cho, E. (2016). Making reliability reliable: A systematic approach to reliability coefficients. Organizational Research Methods, 19(4), 651-682. doi.org/10.1177/1094428116656239
@@mronkko Thank you very much for the commentary and the reference, professor!