Statistical test or method to analyse a change in a variable over time.

#1
Wondering if you can help. I work in the charity sector and I am trying to understand how local government welfare schemes impact the number of people using food banks. I have a data set for 3 years, listing annual local government expenditure on welfare schemes. I also have a data set for the same 3 years showing the number of people using food banks. What I want to know is, how can I check whether an increase in expenditure for a local council is associated with a decrease in the number of people using food banks. Is there a particular statistical method that would be useful here? I am well aware that any change in the 2nd variable could be based on a large number of factors, with the first variable perhaps just one of many. thanks :)
 

Karabiner

TS Contributor
#2
So your study will be based on n=3 observations for annual local government expenditure,
and n=3 corresponding observations for number of people using food banks? Does this
database really require (or justify) sophisticated statistical methods?

With kind regards

Karabiner
 

Miner

TS Contributor
#3
One simple approach that you can try is to plot the time series (# people using food bank) on an individuals control chart. This is a technique developed in industrial statistics to determine whether a process is stable and predictable, or under the influence of outside causes. If your data are stable, then expenditures do not have an impact. If your data are not stable, do the perturbations coincide in time with policy changes (allowing for any time lag in implementation).
 
#4
One simple approach that you can try is to plot the time series (# people using food bank) on an individuals control chart. This is a technique developed in industrial statistics to determine whether a process is stable and predictable, or under the influence of outside causes. If your data are stable, then expenditures do not have an impact. If your data are not stable, do the perturbations coincide in time with policy changes (allowing for any time lag in implementation).
Thanks. I don't know if this would work as I have a set of 152 observations taken at 3 points in time. If i understand the technique correctly, I think I would be taking the mean of these observations and so end up with 3 means. But for this to work, and for me to be able to derive upper and lower control limits, I think I would need more data points. (Apologies if I am not understanding correctly. I don't have a background in quality control so this is all new to me).
 

Miner

TS Contributor
#5
You are correct. I wasn't certain how your data were collected. You can use a 1-way ANOVA to determine whether any changes among the three groups are significant.