# I’m not sure if I’m asking about something really simple or really impossible!

Home Forums Methodspace discussion I’m not sure if I’m asking about something really simple or really impossible!

Viewing 4 posts - 1 through 4 (of 4 total)
• Author
Posts
• #1249
Daniel Guy
Participant

Hi all,

I need to know how to find a correlation between two sets of numbers.The first three numbers are the average percentages of time a zoo animal spent on-show during three different treatments (71.07, 63.38 and 73.66). The second lot of numbers are the percentages of people that saw an animal during the three treatments (81.82, 80.95 and 69.51). I have Minitab 16.0 and have used it successfully for other data in my study but this one has me stumped; can anyone help with the statistical test required to find the presence/lack of correlation? Many many thanks!

#1252
Dave Collingridge
Participant

Basic correlation usually involves two sets of data where the pairs are independent. Are your pairs from the same or different animals?  If the pairs are from different animals, how many animals did you study? Was it 2, 3, or something else?

#1251
Daniel Guy
Participant

Firstly, thanks for the reply. The study was on two animals and each was observed ten times under each condition, so I ended up with 20 durations of time spent on-show per treatment (both animals combined). Using data from both animals combined I found the mean time spent on-show during each treatment  and converted it into percentage of time spent on-show during each treatment (T)  (T1: 71.07%, T2: 63.38% and T3: 73.66%).

During each treatment I carried out a survey asking whether visitors saw an animal, the percentage of respondents that did see an animal in each treatment was found; (T1: 81.82%, T2: 80.95% and T3: 69.51%).

I really appreciate the help, I have uploaded a graph of what I have tried to explain, hopefully it will bring some clarity. It is quite clear from the graph that there is no correlation with on-show time and percentage of visitors that saw an animal but statistical evidence is the order of the day.

#1250
Dave Collingridge
Participant

The possibility of running a traditional test of significance is complicated by small sample size and lack of independence. I recommend a time-series analysis where you plot the time spent on-show per treatment across time. The plot can be divided by vertical lines representing treatments 1, 2, and 3. You can combine the times for both animals or plot the animals separately. If the animals were not supposed to differ in any meaningful way then try averaging their values for each time period. If the animals differed in some meaningful way, plot them individually.

Viewing 4 posts - 1 through 4 (of 4 total)
• You must be logged in to reply to this topic.