Understanding Society User Support

Good afternoon,

I'm hoping to use US to look at the correlations between mood and exercise for different cohorts of the population. I'm using the g_swemwbs_dv variable to represent mood and hoping to use the exercise group of variables (mainly g_vday, g_mday .. g_vwhrs/g_vwmin and g_mwhrs/g_mwmin).

When I look at the data out of the 42,000 odd responses, and 37,099 excluding proxy responses, there is an extremely small sample of people left who recorded doing any exercise whatsoever over the past 7 days. I don't think this can be a realistic representation of the exercise that people really do, and was just wondering if it was something I had done wrong while I was cleaning the data, or whether these questions were not answered particularly accurately by the respondents?

Any help or suggestions would be greatly appreciated because I've come to such a dead end and not sure whether I should continue using this data to answer the question I am addressing about the correlation between mood and exercise participation.

Thank you,

Emma