model

[CMU App Store Study]

Firstly Thank you very much for this kind of good research paper which , I have few questions relates to you research,

 

first: you make this research based on paid apps however ı think the equations will also be current when we equate p to 1  in equation 3 for free apps. right?

second : in equation 13 and 14 you use scale parameters which you got from top 300 apps data howewer the shape parameter in 13 and 14 results from top 200 paid apps,and you used this values in same equation(13,14) is this possible or right?

lastly: On page 21 annotation 14 whether did you estimate "af " value only for top 200 free applications or not and how can we extrapolate for other rank counts easily?

 

SU,

(12/4/2012)

[Rajiv Garg]

 

Thanks for your kind words. I am glad you liked our work.

 

As for your questions:

1.       For free apps the price (p) is 0 so we cannot use it in equation 3 to estimate the distribution for free apps. But we can recover the model by following the steps in the discussion section on free apps.

2.       Yes, both shape and scale parameters are independent of the size of ranked list. They can be used to infer demand for any ranked app as long as that rank is visible. Also, note that the shape parameter usually doesn’t change much with time but scale parameter may change and thus if we calculate the scale parameter today it will be different.

3.       We did estimate the af based on data for top 200 apps, but please note that the estimated value of af is also independent of the size of ranked list. So, if you have a list of 500 apps, the af value will still work.

 

I hope these clarify your concerns.

 

RG

[CMU App Store Study]

As I understood, please confirm me:

 

1-we can use the af value as one which you provided on page 21 annotation 14 in equations 13 and 14 for free app rank lists.(I ask because ı actually dont know how  to solve the estimation problem)

 

2-I actually didnt find aggregate download number where did you get from?

 

3- For example if we know the exact download value of an free app in rank list  we can simply calculate  bp value by using equation

    9 and find daily download.

please excuse me for my too much questions.

 

SU,

(12/4/2012)

[Rajiv Garg]

Yes, you can use the af in the annotation, but you cannot use it in equations in 13 or 14. You need to use the following equation:

Df = bf*rf^(-af)

Where bf needs to be estimated possibly using actual data.

 

The aggregate numbers were provided by Distimo, and I am sure most app analytic firms should have those numbers.

 

You don’t need equation 9 if you have actual data. You can just use equation 2.

 

RG