Difference between self-reported and actual measurements

#1
Hello everyone, I am completing my Dissertation on Social Media usage and Depression.

In the study we were measuring how much time participants were spending on Instagram daily. We had two measurements to record that. One was a self-report question, where participants were simplying reporting how much time they thought they were spending daily on the application. The second measurement was measuring the actual time spent on the application, asking to the participants to get the actual data from their devices (i.e. on the application itself or the screentime on their devices). One of the hypothesis was that the participants would underestimate the self-reported time on Instagram, compared to the actual time spent on it.

Now, I was wondering, which type of test on SPSS I should use to check if the hypothesis is accepted? So basically I would like to see if the participants reported same or different results in the two measurements, and if the results were different, were they underestimated or overestimated?

Thanks a lot to everyone!
 

hlsmith

Not a robit
#2
Post a histogram of reported and actual and difference between then. So three histograms. Also post your sample size. Lastly, do you hypothesize the differences may vary based on any other variables?
 
#3
Post a histogram of reported and actual and difference between then. So three histograms. Also post your sample size. Lastly, do you hypothesize the differences may vary based on any other variables?
Thanks for the reply! The Histograms look quite similar as you can see. First one is the actual data while the second is the self-reported. Unfortunately there is no point to make a Histogram with the difference in scores, as participants were not reporting the exact number of hours/minutes, but rather selecting between 5 options ranging from 10 minutes or less to 2 hours or more.
The histograms just tell me the frequency of people selecting the various answers. So, how do I see for example the difference for each participants, such as see what participant 1 put for self-report and actual data and see if they are the same or different, and so on for every participant? There must be a easier and faster way to do it rather than check one by one.

The sample size is 344 participants

And last, no we did not hypothesised that the difference was caused by any other variables
 

Attachments

#4
Well I was going to tell you to model the differences, but that seems to be out. You will need to look up what agreement stat you need for what I will loosely call ordinal data. It will be a derivative of the Fleischman Kappa inter-rater reliability stat.