02-18-2015 02:36 PM
I have this data sets from couple years for every month(one observation for the entire month) on some different variables related to website usage. I want to see trend , ie. when people are using my website the most and whether any increase in the views into my site as years go by.
I am not coming with a way to solve this issue?
Please help me in figuring how can I look for trends in data.
Thanks in Advance.
02-18-2015 03:26 PM
Your question is extremely broad, and so there really isn't a single answer.
Are you looking for time series trends, seasonal trends, linear trends, non-linear trends, some other type of trend?
02-18-2015 03:32 PM
I have data from my website usage. ie how many people are using my website every month for 7 years. And how many new users are coming every month.I want to design a marketing campaign targeted to improve my website usage by people. So before that I need to know is their a way to know what exactly is the trend of site usage for all these years. I hope i make my query clear this time.
02-18-2015 04:31 PM
My guess is your site usage increased in the past 7 years.
Now what do you plan to do with that information? Question is still too vague to do anything with.
Do you have covariates? For example, the number of users registered, events that occurred in history, conferences related to your field, big news in your field, marketing campaigns? If so, you can look into some time series analysis methods, look into the procs in the ETS package.
02-23-2015 04:22 PM
I am still struggling with this time series analysis. I tried ETS package from sas, but I am not very clearly understanding those procedures without knowing the theory behind.i.e how to choose parameters required for that analysis.
So far we don't have information about any campaigns done to increase the website usage. I just have plain data sets form website usage, and I am looking at trends of website usage, and new users. and little bit about users behavior such as how many average pages users visit every time users come in.I tried OLS regression as well, but model cant explain the good amount of variability, only 42% is explained by model (uni-variate regression).I am stuck at this point, can you suggest any way out.
I have attached a picture of my data for your reference.
02-23-2015 04:39 PM
My suggestion would be to hire a consultant.
Otherwise, post in the Statistics forum and see if others have a response, but I still think the question is too broad for this (and other) forums though you're welcome to try.
Here's some things to consider that probably don't answer you question:
What's your rate of sessions/registerd users? How are the sessions distributed, do a few alumni use most of the services or is it distributed?
What do individual patterns look like, i.e. a user logs in regularly vs occasional.
Also, what's the number of potential alumni.
There looks to be seasonality in your data, can't see clearly if its weekly or daily so you may want to convert your data to weekly instead of daily to smooth things about a bit and re-try a basic regression.
Regarding using OLS, are the assumptions of OLS met in time series data?