Hi
For my dissertation I am looking at air temperature and light intensity effects on water temperature.
I have 16000 induvidual pieces of data (temperature logged over 2 months, every 5 minutes) and when I do a kruskall wallace test, and even a one way ANOVA, all the P-values come out at 0.00, and this is because I have so much data. Only if I cut the amount of data I have down to 50 points so I suddenly get a P-value of 0.02.
Is this an error, or is it because I have so much data, Minitab is comfortably saying it is certain (in this case) that the medians, and the means of what I am calculating are statistically different?
Many thanks
Either all your values are significant or it is indeed a function of a rather large sample. I suggest that you randomly draw 10% from your sample (in equal sized subsets from each 5 minute sampling) and rerun the stats. If you still get the same results, then you can probably assume that the statistics are accurate.
Have you plotted out the values to visually inspect the distributions?
You may consider to estimate average for some fixed time based on the data of the time interval. This will give a precise estimate of the temp as well as reduce the number of data. Then use KW test.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques