Mediation analysis with an ordinal mediator

Question

I have to do mediation analysis for my research and my IV and DV are both continuous and mediator variable is ordinal (time spent on Instagram where categories are:

1. less that 30 minutes a day
2. 30 mins to 3h a day
3. 3h to 6h a day
4. more than 6h

I labeled this mediator as ordinal since measure differences in the categories aren't equal.

So I'm using SPSS and process macro only can be used if the mediator is continuous.

So what method do you suggest? I have been searching all over the internet and no answer found, so I'd really appreciate if you helped...

Answer 1

One option is to assume some distribution within each category and then assign the people in that category a specific number of minutes, then use that as a continuous measure. I would try several assumptions, especially for the lowest and highest category. Use any other knowledge you have.

In a general population, I'm guessing that a lot of the people in category 1 actually spend no time at all on Instagram.

Another is to either not use that macro or not use SPSS. R and SAS both have a method called optimal scaling, which is useful for ordinal variables in regression. They have other scoring options, too. Mediation analysis can be done in a few ways but, AFAIK, all of them rely on regression and no other methods.