It is quite common to do EFA on Likert-scored items, but I would not recommend PCA because of the assumptions it requires. Instead, I personally would prefer Maximum Likelihood using oblique rotation (to allow for correlated factors).

Because there have been so many questions here about Likert-scored items, I have compiled a set of resources on this topic:

https://www.researchgate.net/post/How_can_one_help_people_create_simple_scales_from_Likert-scored_items

Dear David, 

Thank you for your contribution. 

The problem is that after I conduct factor analysis (PCA) I would like to run regression. If  I use oblique rotation with ML I will get correlated factors, which I can't use in linear regression

What do you think about it? 

Best, 

Davit 

Of course you can use correlated variables in regression. That is only a problem if your variables are highly correlated (i.e., multi-colinearity).

Of course you can do Exploratory Factor analysis on 7 point scale.PCA can be used either your indicator is formative or reflective. I suggest you to use EFA (maximum likelihood) if you have more number of responses otherwise you can use PCA.

Best Wishes

Dear Shahbaz, 

Can you suggest any article that suggests to use 7 point scale in EFA. I have seen discussion that 10 point scale can be considered approximately interval (quasi-interval) and it works well in EFA. But when I checked cronbach's alfa for 7 point scale (inter-item correlation) out of 24 items around 10 does not correlate well with other (

The number of points on Likert-scored items is unlikely to have a large effect on correlations, especially at the level of 7- versus 10-point items. Lower correlations are far more likely to be be due to question content and wording.

I, and most others, consider 7 pt. Likert items to represent interval data.  Feel free to apply PCA to the items.

Thank you all for your contributions and informative answers 

It is highly appreciated 

Best regards,

Davit