0
$\begingroup$

I have a collection of 108 data points in the following format:

 page height | % of users who scrolled upto 25% of page | 50% | 75% | 100% (full page)

I'm trying to find the influence of page height on the percent viewership of scroll.

I tried to do regression testing with R, but completely got lost when I started implementing it because the numbers didn't make sense.

My basic formula structure with R was:

data <- lm(page_height ~ twenty_five, data = pagesize_table)

The goal is:

  • How can I correlate the relationship between page height and % scroll?
  • How can I identify the cost of an additional page height versus % scroll? This is tricky because what I want to say is something along the line of "For every 100 pixels added, you lose Y% of users. I know this is possible if I can generate a regression model formula. But not sure how to do it when there are three levels involved.

Head of data:

     V1     V2     V3     V4     V5
1  2318 0.1968 0.3793 0.0519 0.0750
2  3402 0.4859 0.2270 0.0162 0.0619
3  5804 0.0756 0.6321 0.0414 0.0080
4 17431 0.1986 0.2838 0.0039 0.0104
5 11841 0.2969 0.3085 0.0000 0.0012
6 13884 0.3837 0.0384 0.0008 0.0000
Improve this question
$\endgroup$
9
  • $\begingroup$ What is the nature of the twenty_five variable? Is it a column of 1's (person did) & 0's (didn't) scroll 25% of the way down? What does head(pagesize_table) look like? $\endgroup$
    – gung - Reinstate Monica
    Commented Aug 28, 2015 at 2:45
  • $\begingroup$ @gung It's % value and V1 (or page_height) is written in pixel. I added a head output $\endgroup$
    – Adib
    Commented Aug 28, 2015 at 2:58
  • $\begingroup$ So is V1 page_height, & the rest are the"% of users..."? How were those %'s computed? (They don't seem to be cumulative, eg.) What is a row here? Usually, a row is a single observation (eg, a patient). $\endgroup$
    – gung - Reinstate Monica
    Commented Aug 28, 2015 at 3:34
  • $\begingroup$ @gung Each row is a single observation. You can think of it as a patient of different heights, and the percentage represents how much of a medicine traveled through their body (top to bottom) $\endgroup$
    – Adib
    Commented Aug 29, 2015 at 1:38
  • $\begingroup$ What are the different columns, then? V1, etc, isn't very descriptive. $\endgroup$
    – gung - Reinstate Monica
    Commented Aug 29, 2015 at 3:39

1 Answer 1

Reset to default
0
$\begingroup$

You should try using the cor function.

Also your regression formula might be off.

Try: data <- lm(V1 ~ V2+V3+V4+V5, data = pagesize_table)
Also, make sure to use variable selection techniques like backwards eliminate In addition, try glm instead of lm Furthermore, much like the sight, cross validate your final model using another method like forward selection or BIC.

Improve this answer
$\endgroup$
1
  • 2
    $\begingroup$ We are trying to build a permanent repository of high-quality statistical information in the form of questions & answers. We try to avoid one-liners, which would usually be more of a comment than an answer in its own right. Please expand on your answer (for example, you've only answered the first part of the question). If left as it stands it may be converted to a comment. $\endgroup$
    – Glen_b
    Commented Aug 28, 2015 at 2:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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