UCLA Lonliness Scale

#1
I'm writing my first research proposal for my studies, comparing loneliness levels in rural and urban areas

Please can someone clarify if the UCLA Loneliness Scale is a Likert? It only has 4 response categories and I thought a Liker had to have an odd number so there is a neutral?

Also what would be the best way to analyse these results I was thinking of a Pearson’s correlation test to assess whether the overall score is related to living in a rural / urban location
 

timtak

New Member
#2
I am only a learner too so please take the following with a pinch of salt.

I did not know but you are right to say that scales that use even numbers of responses without a midpoint are not Lickert scales. However, they are often called "Lickert-type" scales and I think that the statistics, e.g. of Pearson's correlation are the same or even better on a scale with no midpoint.

The problem with a midpoint is that it may mean mid way between the two extremes (often and rarely in this case) or it may mean "I don't know" or "inappropriate" e.g.
In answer to "My social relationships are superficial" a mid point answer may mean that they are sometimes (i.e. mid way) superficial, or it may mean "I don't have any social relationships so I can't answer" which is in a sense off the rarely end of the scale (but it would not register in that way).

That said there has been research that investigated the mid point and how it is interpreted and found that it was not messing up the stats.
https://journals.sagepub.com/doi/abs/10.2466/pms.1987.64.2.359?journalCode=pmsb

Yes, you could use Pearson's correlation coefficient if you can find a linear numerical measure of rural-urbanness either simply by self report or perhaps better would be first part of post-code / zip code followed by an investigation of population density for a more objective index.

Alternatively you could simply separate the respondents into two rural and urban groups some how (again, self-report of objective measure) and then use a T-test to see which group is more lonely.

I think it is interesting.

The problem for me would be the self-report nature of the scale. My guess is that urbanites are more lonely but they may not self report in that way because, as humans tend to do, they believe that we can do things if we want to, such as from the loneliness scale, subjects may feel they could turn to someone if they wanted to, being surrounded as they are by other people. The only item that directly attempts to measure the "lonely in a crowd " aspect of being an urbanite is the last "People are around me but not with me "

You might consider adding more items like the last or perhaps adapting another similar scale to measure behaviours. E.g. the social support scale is another subjective scale containing items like

If I wanted to go on a trip for a day (for example, to the country or mountains), I would have a hard time finding someone to go with me.

Which might be adapted to behaviors such as

When I last went on a trip for a day (for example, to the country or mountains), I found someone to go with me.

But as I say I am a learner.

Tim
 
#3
A likert scale can have any number of responses, and in fact I was often taught to have an even number when possible, as it enables people to fall less in the neutral area, which is a tempting place and frustrating for analysis.

For simplicity I think a t-test would be the most appropriate. One-ANOVA is a possibility also.