7 things you should avoid in a Data Science Interview

No matter how better prepared you are, data science Interviews are stressful for everyone. Despite having tremendous academic achievements and necessary skills, data science interviews can be very intimidating for most of the candidates. 

Data science is one of the highest paying jobs today. With increasing demand, the number of people opting Data science as a career are massive. However, unlike other fields, data science interviews aren't easy. Despite having relevant skills such as machine learning, deep learning and new-age programming languages like Java, Python and R, candidates face numerous rejections.

According to experts, people often tend to become habitual of putting everything in one box like other technical jobs. In a data science interview, an industry expert said, “since you have plenty of data to compute and come with so many patterns, no matter what approach you have, if you do something over and over, you’ll eventually come up with a result. That’s not what we want.” 

Many experts have shared that they look for clarity of thought. “...with huge data, we don’t want people to keep using the traditional statistical techniques, machines could do that.” Candidates are advised to brush up their basics and keep in mind what “not to speak” during the interview.

Mistakes in Data Science Interviews

1. Logistic Regression is a classification algorithm and “not” a regression model 

Logistic regression is a supervised classification algorithm. It is based on the concept of probability. Logistic regression can be used for classification by introducing a threshold on the probabilities it returns, however that's not the only purpose it serves. It was developed for regression purposes only as the name suggests. 

2. Stepwise regression is a type of regression

Stepwise regression is a step-by-step iterative construction of a regression model that involves selecting independent variables to be used in a final model. It typically involves adding or removing potential explanatory variables in succession and testing for statistical significance after each iteration. Stepwise regression is used in data mining.

Also Read: Top 50 data structure interview questions and answers that tech giants ask!

3. Is the p-value the probability of chance, by chance?

The 'P' stands for probability, and measures how likely it is that any observed difference between groups is due to chance. However, the p-value is affected by the sample size. Larger the sample size, smaller is the p-values. 

4. Confidence interval is the probability that the mean will range between 50 kilograms and 70 kilograms is 95 per cent 

Confidence interval is one the most important topics as it plays a major role in NHST inference. Nevertheless, it is one of the most misunderstood topics as well! here are some common mistakes that people make –

• People interpret a 95% confidence interval as saying if 95% of all of the data values in the population falls within the interval. 
• To say that a 95% confidence interval means 95% of all possible samples fall within the range of the interval. 
• Believing that confidence intervals are the main source of error. While there is a margin of error associated, it is not the only source of error and you should always look beyond that during the statistical analysis. 

What is correct?

Confidence Interval is a range where true value exists. The selection of a confidence level for an interval determines the probability of the confidence interval containing the true parameter value. A 95% confidence level means that 95% of the intervals would include the population parameter.

5. Is linear regression valid if the dependent variables are not normally distributed? 

Linear regression is a commonly used type of predictive analysis. According to Wikipedia, linear regression is a linear approach for modelling the relationship between a scalar response and one or more explanatory variables. Now coming to the above question, is it okay not to log-transform? Well, the answer is yes! 

Why do we transform our dependent variables at all?

When the dependent variables aren’t normally, incompatibility of model errors is likely. In order to make the variables better fit our assumptions and make it more compatible, we transform it.

6. Do we really accept the null hypothesis?

People often don’t pay much attention while they say that they accept the null hypothesis. However, we need to stress upon the fact that we cannot do otherwise! We fail to reject the null hypothesis. 

What’s the difference?

Null hypothesis is never accepted in the first place. We accept them only because we are unable to reject them. So, while using this concept, make sure you correctly put your statement. This shows how sincerely you pay attention to the minute details which ultimately becomes essential for a data scientist.

What next?

These were some of the basic things one should pay attention to. Data science is a generic term. Aspiring data scientists should know that knowing the elementary data science techniques and having a grip on the fundamentals is only the first step. There are various domains such as Statistical Modelling and Experiment Design, deep learning, Natural language processing etc. Different companies and job roles demand different skill sets. People often confess a problem that while preparing for the jobs, they often end up becoming jack of all trades yet master of none. 

What are the primary skills required for Data Science Interview?

Having a strong foundation in SQL, Python, Machine Learning/Deep Learning, Statistics and Distributed Computing. Computer Science fundamentals like Algorithms and Data Structures gives a good understanding of everything when you have to put different blocks together. Once you master the fundamental concepts, you can dive deep into whichever domain you work in. 

Not knowing a technique can get you rejected!

Candidates should ensure that they know all the techniques that are used in processing data. For example, if a candidate doesn’t know the significance of the GIGO concept, then all the analysis is good for nothing. Knowing the statistical techniques is only the beginning of your approach, however, statistical techniques are never in itself enough. If they were, why would we need data scientists? 

Summing up...

At last, no matter how good you are with data science and required skills, it’s all about presenting every bit of it in the best way possible in the interview. The most basic thing that people often mistake is that interviews are a test of knowledge. You don’t have to recite your own resume there. You are called for the interview only because the company could find your resume fit their requirement of skills and qualification. What do interviewers look for is what makes you different from the other aspiring candidates. Experts have said that “clarity of thought” is something that most companies want more than everything else. With large data, people often tend to get lost and lose track of what they actually want to do with it. knowing the techniques are good only if you apply them correctly.