+ Reply to Thread
Results 1 to 2 of 2

Thread: I need some help for my thesis about t-testing different data sets

  1. #1
    Points: 5, Level: 1
    Level completed: 9%, Points required for next Level: 45

    Posts
    1
    Thanks
    0
    Thanked 0 Times in 0 Posts

    I need some help for my thesis about t-testing different data sets




    Hello everybody,

    I'm conducting research on price dispersion to show that offline price dispersion is significantly smaller than online price disperison. I have up until today, monitored and collected over 1500 prices in excel on several homogeneous products in dual-channel/online/brick-and-mortar stores.

    I have showed the price dispersion using standard deviation, variance, Price percentage difference and coefficient of variance. When i showed him the results he said that i needed to show it was significantly higher/lower and that I should perform a test in excel.

    My question is: What test do I need here, excel has three different T-test a Z-test and F-test and I have absolutely no clue how I can compare these:

    What I have done now is averaged the prices on the same days for online shops, and averaged the prices for the same goods on the same days for brick&mortar shops and performed a two sided t-test with uneven variances (since it are different shops). Could this be right and how could i conclude that indeed the price dispersion is lower/higher?

    I really need some help, it would be so much appreciated. Please let me know,

    Kind regards,

    Friso

  2. #2
    Points: 3,506, Level: 37
    Level completed: 4%, Points required for next Level: 144

    Posts
    129
    Thanks
    3
    Thanked 26 Times in 26 Posts

    Re: I need some help for my thesis about t-testing different data sets


    Unpaired t-tests will compare the average price online and offline, not the dispersion. Seems like you need a test for equality of variances. Most statistical packages will run this test for you, but I'm not sure Excel will. The formulas are not too difficult, however.

    One trick, however -- are the dispersions heavily affected by a few outliers? Comparing dispersions with non-normal distributions is more difficult. One simple approach would be to compare using the equality of variance F test, then remove a pre-determined number of high and low values from BOTH samples (say 10 top and bottom from each -- or a constant % like 1% if the counts were different). Rerun.
    Last edited by EdGr; 06-22-2017 at 12:15 PM. Reason: typo

+ Reply to Thread

           




Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats