MockQuestions MockQuestions
Upgrade

Why is mean square error a bad measure of model performance?

1 of 35 Statistician Interview Questions and Answers Written by Rachelle Enns

Updated on July 1st, 2018 | Rachelle is a job search expert, career coach, and headhunter
who helps everyone from students to fortune executives find success in their career.
How to Answer

Assure the interviewer that you are familiar with basic statistical concepts and best practices by briefly walking them through your understanding of why mean square error is not an ideal measure of model performance.

Professional Answer Examples
General
Answer example

"I feel that mean square error relies too heavily on unknown factors which makes it an unreliable measure to many industries."

Written by:

Rachelle Enns
Rachelle Enns is a job search expert, executive headhunter, career catalyst, and interview coach. Utilized by top talent from Fortune companies like Microsoft, General Electric, and Nestle, she helps professionals position themselves in today's competitive digital marketplace. Rachelle founded Renovate My Resume and Executive Resume Solutions, two companies focused on helping job seekers get their edge back. She helps everyone from new graduates looking for their first placement, to CEO's who want more out of their career. Rachelle coaches students to executives on how to master the toughest interview questions and how to handle the most bizarre interview situations; all with confidence and poise. Rachelle trains other career coaches, recruiters, and resume writers, globally. A big part of her job is also spent coaching HR professionals on how to bring the human touch back into their interview and hiring process.
First written on: 03/20/2014
Last modified on: 07/01/2018

View user-submitted Answers

Why is mean square error a bad measure of model performance?
1.
The answers to questions like this usually boil down to how the model is being used. Suppose you are a store owner using a model to predict how many widgets to stock. If you under-predict, you lose the profit on the widgets you could have sold. If you over-predict, you have to deal with the extra widgets on your shelves. Perhaps the widgets are spoiling. The costs of overpredicting may be different from the costs of underpredicting, and these costs are in dollars, so there is no reason to be squaring things. This is an example where squared error may not make sense.
2.
In statistics, the mean squared error (MSE) or mean squared deviation (MSD) of an estimator measures the average of the squares of the errors or deviations, that is, the difference between the estimator and what is estimated. MSE is a risk function, corresponding to the expected value of the squared error loss or quadratic loss. The difference occurs because of randomness or because the estimator doesn't account for information that could produce a more accurate estimate.[1]
3.
One case where MSE will be a bad measure is those cases where linear regression’s assumption are violated. Also, the mean squared error can only be compared between models whose errors are measured in the same units.
4.
When it is under predicted or over predicting.
Get More Interview Practice
Pick your topic
Ask the Interviewer
Questions you may consider asking the interviewer, about ...
Behavioral
We all have some behaviors that are typical of us and whi...
Common Interview Questions
There are some questions that employers ask at almost eve...
Leadership
If you are interviewing for any type of leadership role; ...
Telephone
Most interviews start with a telephone interview. Hiring ...
Tough
Everyone struggles with tough interview questions. Learn ...
View more interview questions