Hi RG

Hope you guys can help!

I am (hypothetically in a paper) redoing the experiment by Fischer et al from 2015) slightly varied to test whether fluency of processing information affects self-assessed feelings of knowledge. That is, e.g. if something is easier to read (more fluently/easily processed), does that affect your cognitive self esteem (as measured e.g. by how able the participants feels in regard to answering a question).

Have questions in terms of modelling the Likert Scale or whether I could restructure the experiment to avoid those issues.

The experiment will be something like this (again, heavily inspired by Fischer et al, 2015, but put here for context):

Induction phase: Participants get questions like "Why do we laugh?" that they need to Google, and fluency (readibility, essentially) of the webpage will be manipulated to be 1-4 levels of increasingly hard to read. I have yet to figure out whether it would be better to split participants up in a between-subject design in 4 groups or to have within-subjects and repeated measures. (Maybe you have thoughts). The increasingly hard levels would probably be done by changing the fonts/colors to become increasingly hard to read with a chrome extension. After reading a question and Googling it, participants have to provide the URL of the "most helpful website" and rate their ability to explain the answer to the question from 1 (very poorly) to 7 (very well) - the Likert Scale.

Self-assessment phase: After the induction, participants will go into the second phase. Here, they will rate their ability to answer questions about knowledge in six domains unrelated to the questions posed in the induction phase. Could be weather, science, history etc., could be "How do tornadoes form?" - and to this question, a participant would then be asked "How well could you answer detailed questions about [topic] similar to these?" on a 1-7 scale. (note that they aren't supposed to answer the question, only rate their ability to do so)

The hypothesis is that experienced ease of processing - so that when it is very easy to read - will lead to higher ratings of self-assessed knowledge for other things, and less ease of processing may lead to lower ratings of self-assessed knowledge.

Now, my questions; and I apologize in advance for not having modelled using categorical variables as outcome with more levels than two -

1. When I have to model these things and have the Likert Scale as my outcome variable, I see I can use ordered logistic regression for this. But I have also seen the debate on using the Likert Scale as a continuous outcome instead, disregarding that it is bounded in the ends.

Can I model it with the ordered logistic regression and have a categorical predictor like fluency (with four levels)? Am I setting myself up here in terms of experimental design to fulfil the assumptions of the ordered logistic reg. modelling?

2. Have seen people that use ratio variables as outcomes, but I do not see how that is not the same issue as above in the modelling.

Would it be better to have a scale from 0-100 where people have to put a bar for the rating and then that would make a ratio variable, or is it still the same issue as above? Could I have ratio as an outcome?