Hi Maya,

the DV has enough scale points with at least semi-interval level that can be used as such. The IV looks a bit shaky so if you want to be hyper correct, make 4 dummies out of it. Much more important are potential control variables that block unobserved confounding effects.

Best,

Holger

Thank you!! That is very helpful, which model would you preferably recommend for this?

Hi Maya Garside , you can use ordinal logistic regression in this case. However, the variables named 'Interpersonal trust' and 'internet use frequency' seem to have a lot of categories. It will be helpful for you if you can suppress the categories.

Best of luck.

Maya,

as the DV can be treated as a normal metric variable, you use a normal OLS regression model. As I said, if you can think about confounding variables in your data set, include them as control (or more flexible: control everything instead of variables that might be affected BY your IV and variables that JUST affect your IV, that is, "instruments").

However, if you can manage to identify instruments, use TSLS instead of OLS or SEM with correlated errors of the IV and DV. An instrument should be identified from a theoretical perspective and has to be conditionally independent of the DV (given the IV).

Best,

Holger

If the DV is Likert I would use ordinal logistic regression because ordinal means ordinal not something else. Best, D. Booth

Dear Maya.

Before providing any details, can you highlight how trust will be or was measured?@maya