I have a set of disease cases in the polygon form as an attribute of each city. There are some 180 cities (polygons) that 2-5 of them recorded more than 300 cases, about 100 of them contain 0-2 cases and the rest recorded 2-20 disease cases. I'm going to evaluate the possible correlation between illness and some environmental factor such as temperature, precipitation, etc.
However, the distribution of the disease data is severely non-normal and violates many statistical methods' assumptions.
Do you have any suggestion in this case?
To assess the correlations in your data set you could use a non-parametric correlation measure like Spearman's rho. Also, if you analyse spatial autocorrelation in your spatial pattern, you should use a Monte Carlo/ramdomization approach to determine you p-values. See e.g. http://pysal.readthedocs.io/en/latest/library/esda/moran.html --> p_rand
Another thing you could use is Poisson regression to determine the strength and direction of influence of your environmental factors on the number of local disease cases.
Cheers,
Jan
You could also try fitting a zero inflated negative binomial regression model (google will provide you plenty of resources). Note that if you find significant spatial autocorrelation then you'd need to deal with it, e.g. through spatial eigenvector filtering.
Good luck,
Levente
Two ways are generally used:
1) Transforming original records to try to make the transformed values as normal distribution;
2) Using non-parametric statistics, including rank correlation (i.e., Spearman)
By the way, if samples are not enough, numerical simulations are necessary to understand the possible uncertainty or significance (i.e., a Monte Carlo simulation or Bootstrap resampling)
Kendall's rank coefficient is another way of handling non-parametric data
You can transform your variables to make them normal. Often variables are lognormal, so you just take logarithms. The base does not matter,, but base 10 is convenient.
You need to be careful with variables that trend because many of these are cumulative. You may need to take first or even second differences.
(See: Cointegration & Granger causality.)
Also Jamal Munshi:
https://www.researchgate.net/publication/306376553_Spurious_Correlations_in_Time_Series_Data_A_Note
Article Spurious Correlations in Time Series Data: A Note
Hi!
You should first check your data for spatial autocorrelation, to determine whether you have a pattern or not. Here are some short definitions of each method you could use, depending on the type of data you've got:
http://www.statsref.com/HTML/index.html?two_dimensional_spatial_autoco.html
Cheers,
Karolina
Thank you all for the answers.
Dear Fredric and Feng
I examined all normalization methods, however unfortunately, the data resist non-normal.
- Badges
- Science topic
- Similar topics
- Geoscience