Image by author
I am a data scientist with a background in computer science.
I am familiar with data structures, object-oriented programming, and database management, having been taught these concepts for 3 years in college.
However, as I entered the field of data science, I noticed a significant skills gap.
I did not have the mathematics and statistics background required for almost every data science position.
I signed up for several online statistics courses, but nothing gave me any satisfaction.
Most of the programs were either really basic and geared towards senior management. Others were detailed and built on background knowledge that I didn’t have.
I spent some time searching the Internet for materials that would support me better understand concepts such as hypothesis testing and confidence intervals.
After interviewing for various data analyst positions, I noticed that most of the questions asked in statistics interviews followed a similar pattern.
In this article, I’ll list the 10 most common statistics questions I’ve encountered in data analyst interviews, along with sample answers to those questions.
Question 1: What is p-value?
Answer: Assuming the null hypothesis is true, the p-value is the probability that we will get a result at least as extreme as the observed one.
P-values are typically calculated to determine whether the result of a statistical test is significant. In elementary terms, a p-value tells us whether there is enough evidence to reject the null hypothesis.
Question 2: Explain the concept of statistical power
Answer: If you were to conduct a statistical test to see whether a particular effect occurs, statistical power is the probability that the test will accurately detect the effect.
Here is a elementary example to explain this:
Let’s say we run an ad for a test group of 100 people and get 80 conversions.
The null hypothesis is that advertising had no effect on the number of conversions. In fact, however, advertising had a significant effect on the number of sales.
Statistical power is the probability that you will accurately reject the null hypothesis and actually detect an effect. Higher statistical power means that the test is better able to detect an effect if one exists.
Question 3: How would you describe confidence intervals to a non-technical person?
Let’s utilize the same example as before, where the ad was shown to a sample of 100 people and resulted in 80 conversions.
Instead of saying the conversion rate is 80%, we would give a range because we don’t know how the real population would behave. In other words, if we were to take an infinite number of samples, how many conversions would we see?
Here is an example of what we could conclude based solely on the data obtained from our sample:
“If we wanted to target this ad to a larger group of people, we are 95% confident that the conversion rate would be between 75% and 88%.”
We utilize this range because we do not know how the entire population will react, and we can only generate an estimate based on our test group, which is only a sample.
Question 4: What is the difference between parametric and nonparametric test?
A parametric test assumes that the data set follows an underlying distribution. The most common assumption when conducting a parametric test is that the data is normally distributed.
Examples of parametric tests include ANOVA, T-test, F-test, and chi-square test.
However, nonparametric tests do not make any assumptions about the distribution of the data set. If the data set is not normally distributed or contains ranks or outliers, it is worth choosing a nonparametric test.
Question 5: What is the difference between covariance and correlation?
Covariance measures the direction of a linear relationship between variables. Correlation measures the strength and direction of that relationship.
Although both correlation and covariance provide similar information about the relationships between features, the main difference between them is the scale.
The correlation ranges from -1 to +1. It is standardized and allows you to easily understand whether there is a positive or negative relationship between the features and how powerful this effect is. On the other hand, the covariance is displayed in the same units as the dependent and independent variables, which can make it a bit more hard to interpret.
Question 6: How to analyze and handle outliers in a dataset?
There are several ways to detect outliers in a dataset.
- Visual methods: Outliers can be visually identified using graphs such as box plots and scatter plots. Points that lie outside the whiskers of a box plot are typically outliers. When using scatter plots, outliers can be detected as points that are far from other data points in the visualization.
- Non-visual methods: One non-visual technique for detecting outliers is the Z-Score. The Z-Score is calculated by subtracting a value from the mean and dividing it by the standard deviation. This tells us how many standard deviations from the mean the value is. Values that are above or below 3 standard deviations from the mean are considered outliers.
Question 7: Distinguish between one-sided and two-sided tests.
A one-sided test checks whether there is a relationship or effect in one direction. For example, after running an ad, you might utilize a one-sided test to check for a positive effect, i.e. an escalate in sales. This is a right-sided test.
A two-sided test examines the possibility of a relationship going in both directions. For example, if a modern teaching style was implemented in all public schools, a two-sided test would assess whether there was a significant escalate or decrease in scores.
Question 8: Given the following scenario, which statistical test would you choose to implement?
An online retailer wants to evaluate the effectiveness of a modern advertising campaign. They collect daily sales data for 30 days before and after the ad goes live. The company wants to determine whether the ad made a significant difference in daily sales.
Options:
A) Chi-square test
B) Paired t-test
C) One-way analysis of variance
d) T-test for independent samples
Answer:To evaluate the effectiveness of a modern advertising campaign, we should utilize a paired t-test.
A paired t-test is used to compare the means of two samples and test whether the difference is statistically significant.
In this case, we are comparing sales before and after the ad, comparing changes in the same data group, which is why we utilize a paired t-test rather than an independent-samples t-test.
Question 9: What is the chi-square test of independence?
The chi-square test of independence is used to examine the relationship between observed and expected outcomes. The null hypothesis (H0) of this test states that any observed difference between characteristics is due to pure chance.
Put simply, this test can support us determine whether the relationship between two categorical variables is coincidental or whether there is a statistically significant relationship between them.
For example, if you want to test whether there is a relationship between gender (male vs. female) and preferred ice cream flavor (vanilla vs. chocolate), you could utilize the chi-square test of independence.
Question 10: Explain the concept of regularization in regression models.
Regularization is a technique for reducing overfitting by adding additional information to it, so that models can better adapt and generalize to datasets they were not trained on.
Two regularization techniques are commonly used in regression: ridge regression and lasso regression.
These are models that slightly change the error equation of the regression model by adding a penalty term to it.
In ridge regression, the penalty term is multiplied by the sum of the squared coefficients. This means that models with larger coefficients are penalized more. In lasso regression, the penalty term is multiplied by the sum of the absolute coefficients.
Although the primary goal of both methods is to reduce the coefficients while minimizing model error, ridge regression penalizes huge coefficients more.
On the other hand, lasso regression applies a constant penalty to each coefficient, which means that in some cases the coefficients may go to zero.
10 Statistics Questions That Will Aid You Succeed in Your Data Science Interview – Next Steps
If you managed to get this far, congratulations!
Now you are well versed in the statistics questions asked in data science interviews.
The next step I recommend is to take an online course that will allow you to refresh this knowledge and apply it in practice.
Here are some resources that I found useful for learning statistics:
The last course can be listened to for free on the edX platform, while the first two sources are YouTube channels that discuss statistics and machine learning in detail.
 
 
Natasha Selvaraj is a self-taught data scientist with a passion for writing. Natassha writes about everything related to data science, a true master of all things data. You can reach her at LinkedIn or check it out Youtube channel.