Dear Scholar Fellows,
I would be interested in your opinion whether an ANOVA test (or Kruskal-Wallis-test) can be applicable in the following situation:
- a) 8 laps are driven on a test track, and the vertical vibrations are sampled at 1024 Hz, the lap times are ca. 40 sec.
- b) from every 1024 point different estimates are derived, such as mean per second, variance per second, skewness per second…these are here „second-statistics”.
- c) the second-statistics are grouped by laps 1, 2, … , 8.
I was considering second-statistics, instead of simply plugging in all individual acceleration values (ca. 40*1024 per lap) into an ANOVA test, because the time domain signal as a whole is highly autocorrelated below 1024 lags.
First, I was looking for hypothesis tests, but I am glad to read your opinion. Do you know any experiment, which tested the mean of a variable, derived from sub-samples?
How can I test or quantify the similarities of second-statistics across laps?
Thank you for your time.
Laszlo
Before Perform any test first you have to check the normality of data.If your data is normal then you will apply Anova to compare the mean for more than two groups and if your data is not Normal then you will have to apply Kruskal wallis test to compare the Median for more than two groups.
Wouldnt a time series analysis be more appropriate instead of "cutting" it intop pieces? I have done EEG analyses, and we would analyze the data as a whole, maybe subdivided into the laps driven in your example. With a time-frequency plot for example you could see changes over the course of one ore more laps. But this is one of many analysis methods.
P.S.: since you specifically said that you are looking for a hypothesis test: which hypothesis do you have and want to investigate??
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA