Statistics problem (math help!).. am I being shaved?
 

Here are my stats for an unnamed program. What is key is here how the conversion ratios have dropped so significantly the last two days (we are talking about the 3rd column not the first). Does anyone know how to use statistics to compare how I was doing from Jan1-25 to my ratios the last two days and see if the conversion rations Im getting could be possibly due to random chance?

diffrent traffic cause i see a drop in traffic, so your default traffic converts less

same exact traffic. only thing that has changed is i got a 'raise' on the 26th, but im thinking they are shaving my leads

I always thought shaving the leads would be easier once they are in the system.

dammit no one knows basic statistics?

depends on where you get your traffic from - google traffic can fluctuate like that

looks to me like you fell off for a good keyword, i would say normal fluctuations can be as much as +-20% a day (but more usually around +-10% a day for me at least) but something like -50% doesn't seem to be just a normal statistical irregularity but probably caused by something

A simple chi-squared test should suffice to get your answer, although you have to keep in mind that things like day of the week might be a confounder here.

1. Today is not over. Wait until tomorrow to have someone calculate this. Usually nights are good for us. So your day is not over yet depending on where you are. 26th could be an outlier.
2. Don't upset that "no one knows basic statistics" unless you think it is something that everyone should know. If that is the case, why do you not know?

No one who knows basic statistics finds statistics enough fun to do it for free :2 cents:

Shave or not, bottom line is: your traffic / balance paid.
If you can make more money with your traffic change programs

You average around 18% Jan 1-25 and the last 2 day's around 10%, that's a loss of around 45%. I don't know, depends on your traffic etc but if It doesn't pick up something has happened either with the site or your traffic.

What happened to half your traffic? Not all traffic is created equally, if the half you are missing is a higher quality source, that explains it right there. Comparing 3800 of your uniques to 1700 uniques is almost apples and oranges.

I don't think he is talking about that convertion. He means the % convertions which doesn't have with the amount of traffic to do.

edit: unless like you say that other half of the traffic is much better quality but It was still at ~18% on Jan 21 and 22 with same traffic is last 2.

Exactly! I paid with blood, seat and tears to go through business school and it's madatory statistics courses. That's not something you go around and do for free or for fun. I normally charge around $1500 to $2000 per day for my services and many are happy to pay.

As well as you?

Ok, lemme see... Hmmm... You were doing pretty good, then they started shaving you.

I can tell this from the fucking comprehensive array of data you've put forth.

Group A is a clear winner (99.9%*)

Group A: 18.19% conversion
Group B: 10.53% conversion

http://www.usereffect.com/split-test-calculator

thanks react
yea i asked my stats friend and he told me i needed to use a 2 sample t-test. (it was something like 99.999628% chance something is up)
i was being shaved, as confirmed by the sponsor.

stats rules!

wait! wait! wait!

You said this all occurred after a "raise"? That actually sounds kind of familiar to me, as in the same thing happened to me. Hit me up if you want to compare notes.

So the question is, are you going to out them? or did I miss the thread already?

My statistics say you're doing fine. Donate to Haiti.

Wouldn't it be nice to have a crystal ball that says "This traffic will convert best at X sponsor, that traffic will convert best at Y sponsor..."

No offense, but the "change sponsors" line doesn't mean anything, unless you can somehow prove that you would do better with a different sponsor, and the only way to know that is to change your traffic and test it out for awhile.... it's a gamble, plain and simple.

:Oh crap:Oh crap:Oh crap

IIRC, a t-test requires that the data be normally distributed. Are the mean conversion ratios on any specific day (say Saturday) much different than the others? If so, the test will probably provide inaccurate results.

But if you already confirmed it, stats are no match for verifiable evidence :)

While confirmation from the sponsor is a pretty clear indication, the 2 sample t-test actually wasn't suitable at all.

Sampling requires randomness to be reliable. By comparing one of the worst days to one of the best days, you eliminate randomness and thus reliability. Pre-selecting known data sets to compare almost never delivers statistically valid results.

Aside from that, you're dealing with categorical variables. So you should have used the chi-squared test instead.

show it to your sponsor, its some money they cannot fuck with your as with any newbie