1
$\begingroup$

I have thousands of product info in a database with category, title, description, make, model, price and store name. I have grouped those products by store names and it seem to span to 4 stores. So thousands of products in those 4 stores.

I want to compare every store's product with rest of stores' products to see if a product in store A is present in Store B or Store C or Store D.

After reading about Ranking, I think it's non-parametric models which I am going to need. But by reading Non-parametric Statistics I found a lot of methods there and I have never used any of them before. I have used very basic stats like mean, median, variance, regression etc. I am looking for your opinion about what non-parametric method will best estimate product comparisons.

$\endgroup$
  • 1
    $\begingroup$ Instead of jumping to technical terms like “non-parametric statistics”, you could perhaps give us a little more detail on what you want to learn. Are make or categories of particular interest? Do you want to compare prices between the stores? $\endgroup$ – Gala Jun 16 '13 at 10:29
  • $\begingroup$ Yes comparing prices between the products of different stores. $\endgroup$ – Umair A. Jun 16 '13 at 10:36
  • 1
    $\begingroup$ Even so, what is the aim, in particular? What is it you're trying to find out? $\endgroup$ – Glen_b -Reinstate Monica Jun 16 '13 at 13:23
1
$\begingroup$

My first advice would be to create some plots to visualize your data. Exactly what plot will depend on what software is available to you, what you are comfortable with, and also on the question you want to address.

To know if stores tend to carry products with different price points, I would just plot the price of the various products side-by-side, a bit like the last three plots on this page. You can get a lot of insight from such plots, not only if the average level looks different but also if there is a lot of overlap between stores or if the distributions differ in other ways (perhaps two stores have a similar distribution on the lower end but only one of them carries a few very expensive products or may some store has a wider range of products etc.)

If you want to know if the same products are sold at different prices in different stores, creating an interesting plot is a little more complicated, you could try to represent the differences between pairs of stores or do something like

enter image description here

but it can become unwieldy with a lot of data.

Of course, if there is little overlap in the range of products, the second approach could suggest that two stores are similar even though one sells mostly expensive luxury goods and the other cheap brands only because the few products they have in common are priced similarly. You need to recognize that those are two different question and interpret the results accordingly.

Regarding inferential statistics, one-way ANOVA or Kruskal–Wallis one-way analysis of variance could be relevant to address the first question and repeated-measures ANOVA or Friedman's test to address the second one but given the way you described your problem, I am not sure that running a test would really be helpful.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.