Monday, May 24, 2021

How not to use the results from a survey

 


 

 

 

 

 

 

 

 

 

 

 

 

 

 

At the SlideModel web site there is an article from April 29, 2021 titled How to Become Great in Public Speaking: Presenting Best Practices. A section titled The Fear of Public Speaking begins with the following paragraph:

 

“Despite the fact that most of us love chatting, far fewer feel comfortable talking in front of larger audiences. In fact 75% of Americans have some level of public speaking anxiety. Many also feel frantic about going on the stage.”

 

Where does this article link? It is to another article posted at the Education Resources Information Center (ERIC) by Raja Farhan in the Journal of Education and Educational Development (June 2017  Volume 4 No. 1 pages 94 to 110) titled Anxiety Level in Students of Public Speaking: Causes and Remedies. But that article has nothing to do with Americans.

 

When evaluating a survey, we need to ask a bunch of questions:

 

When was the survey done?

Who was in the sample? Were they chosen randomly, so results can be generalized to represent a population? Or, was it a “convenience sample” meaning they just surveyed whoever showed up.

How many people were in there?

Are there detailed results or just an ‘executive summary’?

Did everyone answer a question yes or no, or were there some ‘don’t knows’?

How were the genders distributed? What is the margin of error?

 

Farhan’s survey was done on a “convenience sample” of only fifty computer science undergraduate students at a private business school in Karachi, Pakistan. If you wanted to generalize the results, then you would need to take a random sample. Also, the genders for that sample are not indicated.  

 

 


 

 

 

 

 

 

 

 

Results from that survey are shown above via a bar chart. For such a small sample size the margin of error is large – 12.6%. (A typical sample size of a thousand will have a more reasonable margin of error of 2.8%).

 

Also, reportedly 75% said they had a fear of public speaking - which makes little sense. For a sample of fifty there only can be even-numbered percentages spaced 2% apart. So, at face value the two 75% values and the 95% value are wrong. To get 75% we would need two ‘don’t knows’ resulting in 36/48. To get 95% we would need six ‘don’t knows’ resulting in 42/44.

 

The statue in Paris with a facepalm gesture was adapted from this image at Wikimedia Commons.

 


No comments:

Post a Comment