0

I have a blog data set, which has essentially many columns such as:

PublishDate     BlogContent    Views     country
01-01-2018        100          1000        US
01-02-2018        120          1400        US
01-03-2018        50           600         US
01-04-2018        60           500         US
01-05-2018        90           1100        US
01-06-2018        110          1200        US

and so on.

  • PublishDate represents the publish date of a article

  • BlogContent represents number of articles published for the country (for the sake of simplicity i have kept it to US in this)

  • Views represents the number of views that blog post got.

  • Country represents the country for which the blog post was made for

    Now using this data I want to find out that if the blogger produces more content for country US, will there are going to be more views. That means statistically prove the relationship between Number of Content produced vs the Page Views.

I can see from the data that it does have an impact. But how can I statistically prove this?

I tried doing Granger Causality Tests but didnt get the right result. I also did simple correlation, and got a high correlation coefficient. But in stats, correlation is not causation. How do I prove that higher content release for US leads to higher page views for the content statistically?

Thanks.

Community
  • 1
Kshitij Yadav
  • 153
  • 8
  • "didn't get the right result" - what does it mean? What is right result? and if you already know it what are you looking for? – Aksakal Feb 15 '19 at 20:53
  • Granger Causality is [not really causality](https://stats.stackexchange.com/a/144453/7071) in the modern sense of the word (as in I dropped a rock on my foot, which caused me pain). There is also an issue of whether the articles depreciate (news), or if they continue to be of interest in the future, in which case the stock, rather than the flow, on a particular date matters. – dimitriy Dec 11 '20 at 22:53

1 Answers1

2

You don't.

To prove causality, you need the input variable to be determined by factors that are independent of anything connected to the output variable. For instance, having the amount of blog content determined by a die. This is something that requires a controlled experiment. It can't be accomplished merely through statistical analysis.

Acccumulation
  • 3,688
  • 5
  • 11
  • I think this is too strong. I love experiments, but I think there could be some plausibly random natural variation that could be used to answer this question (under some assumptions). – dimitriy Dec 11 '20 at 23:02
  • @Acccumulation; your argument seems me contradictory. You, correctly, conclude that "[causality] can't be accomplished merely through statistical analysis." but you begin your reasoning invoking some independences. Can you elaborate more? – markowitz Dec 12 '20 at 18:26