How do I interpret different results from Unequal Variance test

[#1]

I was testing my data for unequal variance and got different results from the Bartlett, Levene, and Brown-Forsythe. The Bartlett Levene p-value said you can reject the null hypothesis that the variance are the same. The Brown-Forsythe said you cannot reject the null hypothesis. Which test should I believe or is the test inconclusive? 

[#2]

Hello, there are many methods for variance homogeneity test, each method has advantages and disadvantages. For example, LEVENE and Brown-Torsytle can be used when the data does not conform to the normal distribution, while Bartlett requires the data to conform to the normal distribution. The LEVENE method is the first priority in many statistical software.

[#3]

Hi @BuhBuhBuh,

Welcome in the Community !

As @Xinghua mentioned it, any statistical test rely on assumptions on the data used for the test.
Since we don't have access to the data you have used to generate this report and these results to help you more precisely, I would recommend reading the JMP Help dedicated to the different statistical tests used for the Unequal Variances tests : Unequal Variances Reports and Statistical Details for Tests That the Variances Are Equal.

Hope this answer may help you in the meantime,

[#4]

Thanks @Xinghua and @Victor_G the links and the assumptions about normality are helpful. I guess my question was more about which test for variance and when, and is there even an answer to that question? For the data I have I am just curious what to do if tests give conflicting results. Or do you need to decide the test before the analysis. I've attached my data here. 

[#5]

If possible, please increase more data, the data is too little presently.

[#6]

Have you plotted the data yet?

[#7]

To add a bit to @statman 's question...at the least I'd have first plotted each set of observations using the Fit Y by X platform and just stare at the scatter plots for no more than three seconds. If you don't see a large discrepancy in spread across the variables...it's probably not there.