I am a junior doctor getting involved in medical statistics, of which my knowledge is very preliminary.

The problem I face is the following:

I want to analyse the effect of a treatment over time, in a group of patients.

The medication was started at time 0 and I have repeated measurements of a hormone (PTH) and other factors which I want to analyse independently from time 0 every 6 months and up to 5 years.

In the same way I want to analyse the mean dose over time, to see if the patients had an increase in dose. I am not interested to examine how each dose increment affected the results. I just want to examine those factors over time.

Performing a shapiro-wilk test showed that the data are not normally distributed.

Which is the best way to perform this analysis?? Is it Friedman's test? Is it Kruskal-Wallis? Or should I use ANOVA?

Finally, there is a problem with missing values, as only a fraction of patients continued treatment after 3 years. How should I manage this? eg. 50 patients at time 0,6,12,24.. but 18 patients from 36 to 60 months. Should I analyse this subgroup separately?

Thank you for your help, I am grateful. ]]>

I'm having an issue re: the statistical test I should be using in research I am undertaking.

Let's say I have a study population of 300 subjects. Each of these subjects have an individual measure of health, let's say Forced Vital Capacity (FVC) which is a continuous measure. I wish to investigate how the distance from their home and workplace to a pollution source (a continuous measure) influences their FVC. Each subject lives in a separate home, but the subjects are from 5 different workplaces each employing 60 of them.

To investigate distance from the subjects' home to the pollution source I used simple linear regression with 'distance from home to pollution source' as the dependent variable and 'FVC' as the independent variable.

I now wish to do the same for workplaces, but for 'distance from workplace to pollution source', the dependent variable is the same value for the subjects sharing the same workplace e.g. as there are only 5 workplaces there are 5 unique distance measurements (even though they all have unique FVC measurements).

Is this an issue, and if so, how would I go about comparing subjects' 'distance from workplace to a pollution source' to their FVC?

I would like to do so independently of their home as well as including both 'distance from home and distance from workplace to a pollution source' together.

Thanks :) ]]>