*star_border*

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Subscribe
- Printer Friendly Page

# Which statistical model should I choose to process exported data?:

[ Edited ]October 2016 - last edited October 2016

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Email to a Friend
- Report Abuse

Good day!

I exported all data, searched queries, clicks, impressions, sorted by the city and by the date, excell sheet is more than 250 000 rows big. Those are some massive amount of data I would like to process of some kind, predictions to the future, compare cities, days and so on.

Which statistical model should I choose to process exported data? Is somebody here who is using statistical models to process the data? What kind of variables do you compare? How do you predict the future impressions, clicks and total costs?

(as for example, anova, manova, regresion, corelation, montecarlo etc.)

Thank you so much!!

*verified_user*

## Which statistical model should I choose to process exported data?:

October 2016

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Email to a Friend
- Report Abuse

Hi Michal;

This is really an open question with many possibilities.

I would take a different approach. Instead of researching for a user behavior / engagement pattern and statistical modeling, I would first look at the many tools and statistical analysis built into Analytics. This should suffice most of the businesses.

I would start isolating the basic metrics you have an interest in, by using the tools provided by Analytics.

**Moshe,**

**AdWords Top Contributor**,

**Twitter | Linkedin | Community Profile | Ad-Globe**

Did you find any helpful responses or answers to your query? If yes, please mark it as the

**‘Best Answer’**

## Which statistical model should I choose to process exported data?:

October 2016

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Email to a Friend
- Report Abuse

Hi Moshe,

sorry for late reply.

Problem is, I have to use some statistical models (study project) but I found good way is ANOVA, simple and fast at least for now

## Which statistical model should I choose to process exported data?:

October 2016

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Email to a Friend
- Report Abuse

This is just a short note for forum users interested in a math approach: **ANOVA** *(short for Analysis of variance)* is a particular form of statistical hypothesis testing heavily used in the analysis of experimental data. Fore more, pls visit:

***** Analysis of variance *****

https://en.wikipedia.org/wiki/Analysis_of_variance

Lakatos

## Which statistical model should I choose to process exported data?:

[ Edited ]October 2016 - last edited October 2016

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Email to a Friend
- Report Abuse

Thank you Lakatos,

My goal is to process the data by date, locations (cities), clicks and average cost per click.

I have almost 2 months of data that Id like to process in some form of statistical experiments for example, if theres similarities between cities or days (for example saturdays are more popular than tuesdays in regards of impressions and clicks). My ad campaigns were not limited by budget for 2 months, CTR was above 8% for all of the ad subgroups and all the data are aprox. equal and fair regarding statistical approach.

I cant track conversions directly as for e-commerce, ads are for services (home repair and contructions of family houses)

Also, I separated ads by What users want and whole campaing was divided into lots of ad subgroups. I can tell now from city A users prefer product 2 and from city B they prefer product 1 more. There are some interesting differences between cities. Strange, one could say people not far away from each other will prefer similar products, not different products.

Its so interesting that I cannot sleep problem is, time is running and I find myself lost with all kinds of statistical models, variables, options (i use excell addon xlstats)

Also I found some interesting facts.

1. Its not important how much one invest into Ads unless his budget can cover all the demand per month

2. QS can be as low as 1 but still can deliver CTR above 10% and lots of coversions

3. Ad text content is connected with landing page content much more that I thought. My landing pages in the beggining now contain text and all the content on the landing page is only about the true meaning of Ad and this boosted impressions, CTR without higher total cost per subgroup ads.

## Which statistical model should I choose to process exported data?:

[ Edited ]October 2016 - last edited October 2016

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Email to a Friend
- Report Abuse

Michal,

Pure math is off topic in this community therefore I'd only like to make short comments on your last 3 numbered statements.

1. Good point. The daily budget caps your daily spend till you reach profitability. When you do, you'll have your chance to gradually increase daily spend as long as there is more and more online demand for what you offer. Stop doing it when demand seems to be depleted. This strategy sounds quite a bit aggressive, however, it often pays.

2. QS is often times mystified by pros. Focus your attention on your KPMs when appropriate.

3. Correct. That's what we refer to as relevance between ad copy and LP.

Lakatos

## Which statistical model should I choose to process exported data?:

October 2016

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Email to a Friend
- Report Abuse

thanks guys for kind reply! Im so happy to have you both here!

(Will forward this to offtopic forum.)