Producing graphs representing segments from different sample sizes

[Berglind]

Hi there,

I'm producing some graphs based on a set of questionnaires. Some are Likert Scale statements and some are simple yes/no/don't know questions. We are trying to split the answers down by type of respondents: Amenity manager; productive managers; multi-functional managers; and conservation managers. We have 3 conservation respondents and 36 productive respondents. So in our graphs, conservation managers are always under represented. I then calculated the percentage within each segment which responded yes/no/don't know (24% of productive managers responded yes, 13% responded no etc - and not all participants responded to all questions, so in one case, only 1 of 3 conservation managers responded, so 33% of all conservation managers or 100% of conservation managers that responded. I've visualised this in simple bar-charts. The danger in this is that it gets quite confusing for people to understand what I've done, and it doesn't reflect the low sample sizes for some of the segments, inflating or deflating the results disproportionately. Is there any better way of doing this?

 

[oldbrewer]

I think I would group conservation managers with amenity managers

 

[Berglind]

Not an option I'm afraid. They have very different views and attitudes, and the segmentation was developed based on outcomes from past studies.

 

[oldbrewer]

You obviously understand low sample sizes mean dodgy information. So I would say only 1 out of 3 responded and no meaningful conclusion can be drawn. And I would contact the 3 and tell them it is imperative they respond.