However, because of their specific characteristics—for
example, respondents having no person in their support networks and/or giving invalid (i.e., item nonresponse) answers
on selected variables (e.g., household income, social class), which were most often caused by explicit refusal or ‘‘don’t know’’
answers—we were left with 91 eligible mobile phone users who provided valid answers on all variables entered in regression
analysis for emotional support and 89 eligible units in regression analysis for social companionship