Dear Statalist,
I have a question relating to longitudinal data. I have a dataset that comprises repeated observations of firms over time (30 firms, 44 quarterly observations each). The variables are aggregated scores from consumer surveys, key ones being:
Customer satisfaction: (% satisfied minus % dissatisfied, neutral scored as zero)
Customer willing to recommend: % saying they will recommend the firm
Non-customers receiving recommendations: % saying this
Non-customers favourable impression: % saying they have a positive impression of the firm (positive minus negative %, neutral scored as zero).
I hypothesise that each one has an influence (i’.e. causes’) the next. In other words:
Customer satisfaction scores for firms are positively associated with willingness to recommend scores
Firm's willingness to recommend scores are positively associated with scores among non-customers for getting recommendations
Firm's scores among non-customers for getting recommendations about them are associated with positive impressions of those firms.
However, I am not sure what the most appropriate analysis would be. I could run a series of separate regressions, but that seems clunky. I thought some sort of SEM since several of the variables are both dependent and independent. But an SEM with many (44) over-time observations on each entity? I have not seen anything like that before.
Attached is an excerpt of the data in excel which may help. I would be very grateful for any suggestions/advice !
John D
I have a question relating to longitudinal data. I have a dataset that comprises repeated observations of firms over time (30 firms, 44 quarterly observations each). The variables are aggregated scores from consumer surveys, key ones being:
Customer satisfaction: (% satisfied minus % dissatisfied, neutral scored as zero)
Customer willing to recommend: % saying they will recommend the firm
Non-customers receiving recommendations: % saying this
Non-customers favourable impression: % saying they have a positive impression of the firm (positive minus negative %, neutral scored as zero).
I hypothesise that each one has an influence (i’.e. causes’) the next. In other words:
Customer satisfaction scores for firms are positively associated with willingness to recommend scores
Firm's willingness to recommend scores are positively associated with scores among non-customers for getting recommendations
Firm's scores among non-customers for getting recommendations about them are associated with positive impressions of those firms.
However, I am not sure what the most appropriate analysis would be. I could run a series of separate regressions, but that seems clunky. I thought some sort of SEM since several of the variables are both dependent and independent. But an SEM with many (44) over-time observations on each entity? I have not seen anything like that before.
Attached is an excerpt of the data in excel which may help. I would be very grateful for any suggestions/advice !
John D
Comment