Can we asses the outliers in a list of frequencies with something similar to Q-Tests ?
I am working on a problem involving the comparison of different target's behavior, and I have noob questions regarding the seriousness of some processes.
Let's say that we have between 4 to 10 targets that are picking one location in a list, and that I'm logging these locations and grouping them by country, and they do it from 40 to 800 times depending on the target. So for each one I have a contingency table of the number of times they hit countries. I am able after some processing (grouping) to have some data ready for a chi square test of independence by fitting this criteria:
>No more than 20% of the expected counts are less than 5 and all individual expected counts are 1 or greater (Yates, Moore & McCabe, 1999, p. 734)
So I can asses the probability to have statistically correlated series but that's not enough.
I want to take for all targets the frequencies of choice for one Country, let's say Japan, and to asses if target X frequency of choice for this country is an outlier regarding the other ones using Q test.
The problem is, not all frequencies are based on the same number of iteration. In fact, if I have a very low cardinality for one target, the incertitude on the frequency will be so fuckin high that my Q test will be unreliable.
Is there any statistical tool that can help me look for outliers in the targets frequency of choice for one country, while taking their number of choice cardinality into account ? Am I going in the wrong direction to asses this ?
Thank you for reading this.
I am working on a problem involving the comparison of different target's behavior, and I have noob questions regarding the seriousness of some processes.
Let's say that we have between 4 to 10 targets that are picking one location in a list, and that I'm logging these locations and grouping them by country, and they do it from 40 to 800 times depending on the target. So for each one I have a contingency table of the number of times they hit countries. I am able after some processing (grouping) to have some data ready for a chi square test of independence by fitting this criteria:
>No more than 20% of the expected counts are less than 5 and all individual expected counts are 1 or greater (Yates, Moore & McCabe, 1999, p. 734)
So I can asses the probability to have statistically correlated series but that's not enough.
I want to take for all targets the frequencies of choice for one Country, let's say Japan, and to asses if target X frequency of choice for this country is an outlier regarding the other ones using Q test.
The problem is, not all frequencies are based on the same number of iteration. In fact, if I have a very low cardinality for one target, the incertitude on the frequency will be so fuckin high that my Q test will be unreliable.
Is there any statistical tool that can help me look for outliers in the targets frequency of choice for one country, while taking their number of choice cardinality into account ? Am I going in the wrong direction to asses this ?
Thank you for reading this.
