What statistical method/formula can be used for Range and Trend?

I have a set of data (attached) which shows the volume of web traffic by Country, trended weekly. I'm trying to learn/find a statistical method that help me analyze the data in the following way:

  1. How much does it change? (Max, Min) -- I actually don't need an answer to this, I know how to do it, just reference for the second question below
  2. What is it's normal range? (Normal highs and lows, not maximums/minimums)
  3. What is the trend? (increasing, decreasing)

Are there statistical formulas to help understand this?
Elizabeth "Smalls" EckelsDigital Analytics ConsultantAsked:
Who is Participating?
 
IanStatisticianCommented:
Hi there Elizabeth,

Lookong at your data I would make the following points -

  • You have 10 time series (one for each country) and for **correct** analysis you would need to use time series techniques**
  • They only show 18 weeks of data.  Is that all there is at the moment?  Can you go back in history to get previous values?  Will this project be going forward in time?
  • Is there likely to be more countries? and would you be interested in inter-country comparisions?
  • It is possible that a trend is going to be definitionally difficult to specify.  Are there likely to be ups and downs related to any of:  Christmas commerce, Summer holidays, political events, ...?  If so are you looking for the trend leading up (potential relative increase) and eventual settling down after (relative decrease) or do you want the trend going through such bumps?
  • Do you just want to describe the data?  Or do you want to use the data to estimate values of the bigger population or the process that generates the values? Or do you want to predict future values or ranges of values?  Your data collection technique/methods will be an important aspect to asddress these questions.
  • Typically with such time based data where it is measured over a longer period (ie at least a few years), it is **seasonally adjusted** with the aim to remove some of the usual up and down swings.  Is that going to be appropriate for you?  - cant be done for 18 weeks; you may be more interested to the seasonal effect rather than week to week change;  These are questions you need to ponder.

** - With data measured through time the inter measurement period is an important parameter.  When measured at points close in time a data point is not giving us a lot of new information  as it is going to be close in value to the previous value.  This concept is measured by auto-correlation.  When you use basic statistical summaries typically mean values will be correct but variances, standard deviations and standard error of estimates will be incorrect (and usually too small) .  Time series techniques (which are more involved) overcome this problem.
 

Hope this helps to clear your mind as to what you want/need.  Get back with more thoughts.

Regards,

ian
1
 
xtermieCommented:
Can't see the attachment
1
 
Elizabeth "Smalls" EckelsDigital Analytics ConsultantAuthor Commented:
0
 
Elizabeth "Smalls" EckelsDigital Analytics ConsultantAuthor Commented:
This was really helpful in my thought process, Ian, and it stunned me in a good way. I realized I don't have enough timespan of data to get a reliable trend yet, so I just took a look at the slope over the time period I had. I might be getting back to you with a question down the road though now that I know what I need as inputs :) thanks so much!
1
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.