Biostatistics student here hoping for some helpful input.
I am working on an assignment, and have a dataset containing summed disease case counts for a country - one case count per year. I've been asked to use the total case count for each year as my outcome variable. The assignment is asking me to test whether there are statistically significant differences in case counts between a set of six different years (2005 - 2010), and the assignment states that total case count should be treated as a continuous variable.
Normally if I were comparing continuous outcomes between six groups, an ANOVA would be my go to. However, I don't have means for these various "groups" (years) - just the total counts for each year. Is there a statistical test that will allow me to test whether there are significant differences between yearly case counts? I'm stuck, currently working in R, and would appreciate any thoughts!
You can try a chi square test if you can work out an expected value :)
Case count is exact so all differences are exact and hence any difference not equal to zero is a real difference. If your instructor means something else he should have written a better question. Best wishes David Booth
Let me try to answer your question by taking a fictious data. The following table shows the year and the disease count. The baseline figure is 15, so for my understanding the expected value for subsequent years would be 15. Assuming this perform the chi-square test for each year.
year 1 2 3 4 5 6 Total
Count 15 17 21 18 24 28
Chi-square 0 0.27 2.4 0.6 5.4 11.27 19.94
You can observe that for the first 4 years there seems to be no significant change. However, for the 5th and 6th year, there is a significant change. So, for individual comparison sake, you can say that the 5th and 6th year showed the significant changes. Overall, also, going by the chi-square of 19.94 for 5 degrees of freedom, it is significant. So, you can say that over the six year period, there is a significant rise in the incidence of the disease cases.
I can also perform the correlation between the year (X) and counts (Y). If the correlation so obtained (0.915; d.f. = 4) is significant, then you can conclude that there is a significant linear rise in the counts of the disease. The slope (b)value for the above data is 2.37 and it has to be taken as significant as the r is significant.
The interpretation of slope (b=2.37) will be that there is a rise of 2.37 cases, every year.
Based on my experience, I would have answered as above.
Taking year as 2005,2006....2010 will not change the results and interpretations of the above fictious data.
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking