What kind of tests i can do to see if two diseases are related

x509

New Member
#1
I am using data from this website (data is at the bottom of the page) for a case-study. For the sake of question lets stick to any of FluWatch sheet say FluWatch-Alberta. I am coming up with a question of my own, i.e., if one disease has any relation with another one. What kind of statistical technique should I use here?

What i have thought so far

  • The data is nominal so correlation wouldn't work
  • Should I do some sort of cross classified observation? (I just read it somewhere not sure what it is)
  • Or I am on wrong track and I can just do correlation/regression to check if disease A is related with disease B

I understand that this data might not be enough but thats all i have. I would like to ignore any sort of seasonal variances and do some basic tests to see if two diseases are related. What can i do
 

hlsmith

Omega Contributor
#2
Can you provide a screenshot of data, to give us an idea. So these are time series data that you don't want to plot using time series? Do these data just have two incidence values or what are you working with, so that I don't have to chase links and open files.


Thanks.
 

x509

New Member
#3
Can you provide a screenshot of data, to give us an idea. So these are time series data that you don't want to plot using time series? Do these data just have two incidence values or what are you working with, so that I don't have to chase links and open files.


Thanks.
The data is already on the website that i shared but here is the screen shot.

https://www.dropbox.com/s/pscpdmr5d1tqab4/Screenshot 2016-11-22 11.42.58.jpg?dl=0
Certain diseases have data during certain time of the year.
 

hlsmith

Omega Contributor
#4
So you are looking at rates in suspected cases, correct? Test results are likely not in the same people, but are general counts, correct? If so, you can say they occur at the same period but who knows if they are actually related or if a person eventually tests positive at a later date.


What is the purpose of the problem - for class or fun?
 

x509

New Member
#5
So you are looking at rates in suspected cases, correct? Test results are likely not in the same people, but are general counts, correct? If so, you can say they occur at the same period but who knows if they are actually related or if a person eventually tests positive at a later date.


What is the purpose of the problem - for class or fun?
its a class case-study and I am trying to get into right direction. not that i am asking someone to solve the problem for me just guidance. Here is what i have done so far

  • I realized that the data isn't nominal. I read this https://statistics.laerd.com/statistical-guides/types-of-variable.php to come to this conclusion.
  • Now i am trying to study if FluPos (influenza) has any effect on RSV diseases. I did t-test which told me to reject my null hypothesis since my t-stat > t-crit (two tail).
  • Then i tried with hypothesis that influenze in general affects all respiratory diseases. I did Anova here and it came out that my F > F crit. So i have to reject null hypothesis here too

For above tests i did only +ve tests of Influenze with pos tests of other diseases. I even tried testinging for the time period where i have data FOR ALL DISEASES.

I am not sure if i am on right track or not. Now, I am thinking just to do simple correlation. I know i don't have much data but i want to do as much as i could with given data.


Update
So i did the correlation between all diseases and i see those do correlate

https://www.dropbox.com/s/frg6llyjugxxtmy/Screenshot 2016-11-22 13.59.18.jpg?dl=0

and here is its chart

https://www.dropbox.com/s/h95e1k152x071rj/Screenshot 2016-11-22 14.00.13.jpg?dl=0


However, I do understand that the data we have can't be used to work on hypothesis if disease A causes disease B because i think that we don't have data that if patient X who got disease A is also got disease B after sometime. The data provides just the number which i believe isn't enough for us.
 
Last edited:

hlsmith

Omega Contributor
#6
Step one, describe the type of data you have, predictor (independent variable) and outcome (Y). Next define the study design. These two things will completely narrow down what type of analyses you can perform.