Consulting required: Transformation or other model?

MPB_MPB_ Member Posts: 45 Guru
Hello everyone,

I am examining a dataset of some sales records. According to the Literature in this Field, there are some delayed effects of pricechanges.
To be more specific, the theory says: If Prices are high(er), consumers are more likely to wait and shift there purchasing actions into the future.
Therefore, I calculated the Price of the previous week for every row in the dataset. Following the theory, I would expect a positive sign for the coefficient "price previous week (-->X-axis) ; (Units --> Y-axis)" in a regression model and therefore, a rising regression Line



As you can see in the graph , I am not even close to such results.

I then thought, It would be a good Idea to transform the data and applied logarithms on both factors with the following result :





I can now see the effect, but only in parts of the whole plot. In a calculation for the whole dataset, there are still negativ signs.

Now to my actual questions:


Do you have an Idea for how I could transform the data so that I can confirm the theory from the literature?

Or is there even a better model for that than regression?


Please excuse the long text and have a great day







Best Answer

Answers

  • MPB_MPB_ Member Posts: 45 Guru
    Hi @sgenzer,

    thank you very much for taking care.

    Actually, you can close that question. I worked around this issue.

    Best Regards
  • sgenzersgenzer Administrator, Moderator, Employee, RapidMiner Certified Analyst, Community Manager, Member, University Professor, PM Moderator Posts: 2,959 Community Manager
    @MPB_ no problem. Thanks for responding.

    Happy RapidMining!

    Scott
Sign In or Register to comment.