In A/B testing metadata, I assume it's currently set to some sort of t-test, but I was wondering IF someone also selects to test their thumbnail (as it's an option when you select a metadata test), shouldn't the test method be ANOVA?
-
Guest - Earn a FREE TubeBuddy Upgrade for being active on the forums! Click Here to learn how you earn free upgrades for TubeBuddy!
Question TubeBuddy Logistics: ANOVA instead of t-test?
- Thread starter Melanie
- Start date
t-test is typically fine with two compared elements. ANOVA is more effective with 3 or more elements. I don't speak for TubeBuddy, but I suspect they are using t-test because it is two thumbnails versus comparing the individual elements of each thumbnail.
OP
OP
Yes, but when you start including other variables like metadata, should that go in ANOVA because original metadata and new metadata are two additional variables on top of thumbnails (which can be added within the metadata test)? Now if someone chooses just to test metadata and not also thumbnails, a t-test would be sufficient. I'm just curious why the metadata change also allows the uploading of thumbnails unless an ANOVA is done.
I think you are assuming there is more sophistication to the A/B test than there is. Again, I'm not TubeBuddy, but it looks like they are just doing a pure CTR analysis.
BTW, I'm not disagreeing with your premise. I'm just not sure the A/B test is that sophisticated. In the output report it says it is looking at CTR. I think the rest of the report isn't via analysis, but rather just reporting.