It seems to me that using a linear or non-linear regression model is appropriate.

You can use regression analysis and correlation analysis. This will help you find the individual relationships

Your sample size is pretty small to test this sort of model. You can run the analysis, but the estimates of your statistics is likely to be very inaccurate estimates of underlying population parameters. My advice is to try to get a larger sample so that your results would be more precise, and what you find is more likely to replicate in another sample.

Paul E. Spector thank you for your response! I know, it's a small sample but:

- there have been some other studies with small sample

- it's been pretty hard to find participants, cause it's a specific app which students have to download and it is on voluntary basis. I've been trying since February. Big sample studies would usually test either non-voluntary technology, or something which everyone uses (ie mobile internet)

Do you think there is something I can do in my situation to improve my statistics data?

Thank you Abolfazl Ghoodjani and Eldard Mukasa for your advice. What do you think, would data from 50 participants be sufficient for regression and/or correlations analyses?

Yes, I think it's enough

Partial least square method smart pls

Since it is master thesis 50 sample size is fine. The most appropriate technique in this case is partial least square Since you are using two dependent and multiple independent variables. Simple correlation and regression analysis will not be useful for the given model...

variance or covariance based structural equation modeling

Path modelling can be performed but data should be linear and normal. In PLS even non linear and non normal data can be used. That is why it is better. Structural equation modelling is preferred when there are both observed and unobserved constructs.

I am totally agree with David.

For PLS-SEM, minimum sample size should be higher than 10 times of the maximum number of inner or outer model paths directed at any latent variable in the model (Hair et. al, 2011). Your sample size should be 70 (7 paths*10), given, all constructs has less than 7 items. If some constructs has more than 7 items (e.g. 8 items), your samples will be 8*10 = 80.

Another choice, you may use Technology Acceptance Model (TAM). Since, both of TAM and UTAUT aim to explain technology adoption. But, TAM use only perceived usefulness (similar to performance expectancy) and perceived ease of use (similar to effort expectancy) in determining technology adoption. Therefore, sample size is lower than UTAUT.

Reference:

Hair, J.F., Ringle, C.M. & Sarstedt, M. (2011) PLS‐SEM: indeed a silver bullet. The Journal of Marketing Theory and Practice, 19(2), 139–152.

I Agree With David Campbell

Good @tanikan

Hi, I have used SmartPLS3 to validate an simplified UTAUT2 Model (7 variables to explain behaviour intention, no moderators) for my masters thesis. I have actually analysed the model already with SPSS and just tried PLS out of curiosity. The model could not be validated (in both PLS and linear regression) but I'd be VERY curious to have someone help me read all the data. Either by Email or even a short VC. Depending on your efforts I'd pay up to €50 for the "coaching" (or donate it to your preferred institution). Would be great, thanks.

Hey Tobi Prade ,

Im struggling with the same Model ins SmartPLS do you have any advice, and could you maybe tell me how you managed it?

In my model, the individual constructs have values that are far too high (Cronbach's alpha, rho_A, etc.).

Effort expectation, for example, has a value of 0.98.

It would be amazing if you could help me!