ANOVA weighted means?

[piblet]

Hi Chaps,

I am trying to run a comparison between two groups - patients and staff - ratings of the quality of life on a ward.

Simple enough, but the answers come from five different wards, and unfortunately the number of staff and number of patients recruited from each ward is not equal.

If I just do an ANOVA or Independent Samples t-test comparing Quality of Life between staff and patients, I will get biased results because there are a higher portion of staff from a ward which would be expected to be 'nicer'.

Is there a way I can adjust for this? I've been looking into using an ANCOVA but note I need a continuous variable rather than a categorical one.

Any clues greatfully appreciated. I've got access to both SPSS and STATA.

Thanks

 

[abnormal distribution]

What is your research question? Is it that the ratio of staff to patients impacts the quality of life? If so, then perhaps you need to rethink your groups. If not, please give some more details on what you are examining and why.

 

[piblet]

Hi, thanks for the reply.

No - it's not to do with the ratio. The research question is simply that staff as a whole will rate quality of life more favourably than patients.

 

[abnormal distribution]

If you're only interested in 'the whole' then I would just pool all your results together into the two groups and be done.

If you want to dig deeper and look at the effect of ratio (which I suspect you do since you were interested in weighting things), then you should probably break the groups down further. Ie, you could create a few ratio 'bins', low/med/high ratio. You could then have groups low-ratio-patients, low-ratio-staff, and so on. Or you could perform a two-way ANOVA and use the ratio (low/med/high) as your second factor. (edit: your second factor could quite simply be ward number)

Btw, I don't claim to be an expert... that's just my take on it.