Hey Hassan Azganin

please provide more information on how you ran the factor analysis (EFA vs. PCA, method of factor retention, method of rotation, factor loadings, output). If seven variables is all you have and you expect them to load onto one (?) factor, I do not see the advantages of EFA compared to CFA - usually I am a strong advocate for EFA.

Best

Marcel

Marcel Grieger Thanks a lot for the prompt response, I tried different options to improve the result but failed, this includes different methods (i.e principle component, max likelihood etc. All rotation methods were used one by one, for the absolute value under the coefficient display format, I tried 0.3; 0.4 and 0.5).

Hi Hassan Azganin

you are welcome. I take it you have not yet tried CFA, am I correct? Why not try it and post the results of the fit indices here?

Best of luck

Marcel

Yes, I did run the CFA analysis of 7 variables, and found an issue with discriminant validity results. Hence, to improve it I deleted items of factor loading more 0.7. for two variables. To illustrate, find the explanation below (check attached files also):

Perceived Usefulness (PU) has eight items all have good loading factor, and perceived ease of use (PEOU) variable has 6 items and all factor loadings are also above the accepted value (>6).

To improve discriminant validity result, I needed to delete:

1) 4 items of perceived usefulness with factor loadings range between 0.76 and 0.77

2) 2 items of PEOU with factor loadings of 0.75 and 0.78

I would like if there is any justification for omitting these items noting that there is no cross-loading issue when EFA was run. The only issue I had with the pattern matrix result is that PU and PEOU are loading under one factor

As per your CFA model fits, CFI value is lower than 0.95 and RMSEA value is higher than .05. If I am not wrong, you have utilized ML estimation. I suggest you to utilize DWLS estimation and check whether model fits improved as well as factor loadings, co-variances, etc. .

There may be nothing "wrong" with your data that needs "fixing". It may be reality.