Talking More About Statistics

Here’s the breakdown of how interested people are in my report on The Great Big Hyper-V Survey of 2011, based on the registered nationality of IP addresses downloading the PDF (I deleted the really low percentages):

Country

Percentage

United States

33.80%

United Kingdom

12.20%

Netherlands

10.45%

Ireland

7.67%

Australia

3.83%

Germany

2.79%

Canada

2.79%

Sweden

2.44%

Japan

2.09%

France

2.09%

Italy

1.74%

India

1.74%

Belgium

1.74%

Greece

1.39%

Spain

1.05%

Norway

1.05%

Many of those USA IP addresses have an address in Redmond Smile

I have received a few queries on some statistical discrepancies between my report and the commentary by Hans Vredevoort.  Kwik Surveys incorrectly (in my opinion) calculates percentages for any multiple choice questions. 

For example, question 46.  Hans chose to share the results of all 614 people that answered questions (including 14 who didn’t use Hyper-V at work and 192 that existed the survey half way through), preferring to analyse individual questions based on as many people as possible.  Approximately 440 people would probably have answered this question, and 309 stated that they use VMM.  The Kwik Surveys report showed the percentage as 29.4%, not the approximate 70% that 309 of 440 actually is.  What Kwik Surveys did was total up all selected answers, 1051 (because around 600 people selected multiple answers, as we requested), and said that 309 of 1051 was 29.4%.  Kwik Surveys calculated percentage badly under represents what percentage of the sample actually use VMM.

image

So that’s why I spent a good bit of time creating a spread sheet (in my raw data download) that takes the number of answer selections and divides them by the number of respondents to get a percentage.  In my case, I only used the answers from those respondents that answered all questions – this allows me to combine and use answers from different questions to draw conclusions.  I also excluded those who don’t use Hyper-V at work – I was interested in cloud and Systems Management topics which would have been irrelevant to them.  In the case of question 46, 282 of 408 respondents said that they used VMM.  That’s 69.12% of the sample.

System Center Essentials

10.05%

41

Virtual Machine Manager

69.12%

282

Operations Manager

43.38%

177

Configuration Manager

30.39%

124

Data Protection Manager

41.18%

168

Service Manager

6.37%

26

Opalis

8.58%

35

Not applicable

20.83%

85

To be fair, I don’t have a life, and I spent a lot of time in the two nights after closing the survey, taking the numbers and calculating the accurate percentages.  Hans, who does have a life, trusted the Kwik Survey report.  And this is our first time doing this.  We’re not professional pollsters.  But we’re learning … and we will get it right for The Great Big Hyper-V Survey of 2012.

My 69.12% report is an accurate reflection of VMM usage, as opposed to the very misleading 29.4% that Kwik Surveys has calculated.

One thought on “Talking More About Statistics”

  1. Thanks Aidan for clearing this up. I indeed have a life and a busy one too. While have a great time in my new job, cooking for three teenaged boys with a wife in the south of France, I managed to find two hours to process the data and finally to write it all up between at 2am 🙂 Before you start thinking I am complaining, I took a plane to Carcassonne last Thursday so that was quite a welcome little holiday. When I get back I will get rid od the wrong counts and conclusions.
    Finally I would like to compliment you on all the work you did in making this survey possible and allowing me to participate in it. We broke every possible record on our hyper-v.nu blog.

    Thanks,

    Hans Vredevoort

Leave a Reply to Hans Vredevoort Cancel reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.