How do I approach Coding & Algorithms interview questions?

Coding & Algorithms questions require understanding of core concepts and practice. PracHub provides solutions with explanations to help you master coding & algorithms interviews.

What difficulty level is this interview question?

This is a easy difficulty Coding & Algorithms question, commonly asked during Take-home Project rounds at Roblox.

What role is this question designed for?

This question is commonly asked for Data Scientist candidates at Roblox during technical interviews.

Implement four DS coding tasks | Roblox Interview Question

Q: Implement four DS coding tasks

This multi-part question evaluates a data scientist's competencies in statistical inference (sample size and z-test calculations), causal inference and parallel-trends validation (difference-in-differences), Bayesian probability updating, and interpretable supervised learning feature importance, all framed as coding tasks.

You are completing a CodeSignal-style assessment (Python or R). Implement solutions for the following four independent questions.

1) Two-sample z-test: required sample size

You are given:

x : numeric array of historical observations for the metric (use it to estimate the metric standard deviation sigma )
alpha : significance level (e.g., 0.05)
power : desired power (e.g., 0.8)
effect_size : the minimum detectable absolute difference in means, $\Delta$

Assumptions:

Two-sided two-sample z-test for a difference in means.
Treatment and control have equal sample size $n$ .
Use $\hat\sigma = \text{std}(x)$ as the population standard deviation estimate.

Task:

Return the minimum integer per-group sample size n required to detect effect_size at level alpha with power .

2) Difference-in-Differences (DiD) + parallel-trend validation

You are given three equal-length arrays:

period[i] : time indicator (contains at least a “pre” and a “post” period; may contain multiple pre periods)
group[i] : 0 = control, 1 = treatment
outcome[i] : numeric outcome

And a numeric threshold for trend validation.

Definitions:

Let $\bar{Y}_{g,t}$ be the mean outcome for group $g\in\{0,1\}$ in period $t$ .
The DiD estimate is:

$\text{DiD} = (\bar{Y}_{1,post}-\bar{Y}_{1,pre}) - (\bar{Y}_{0,post}-\bar{Y}_{0,pre}).$

Parallel-trend / trend validation requirement:

If there are multiple pre periods , compute the group difference $d_t = \bar{Y}_{1,t} - \bar{Y}_{0,t}$ for each pre period $t$ , sort pre periods by time, and validate:

$\max_t |d_{t} - d_{t-1}| \le \text{threshold}.$

If there is only a single pre period, treat trend validation as passing.

Task:

Return (a) the DiD estimate and (b) whether the pre-trend validation passes under the threshold .

3) Bayes’ rule posterior probability

You are given probabilities (as floats) describing an event $A$ and evidence $B$ , such as:

p_A = $P(A)$
p_B_given_A = $P(B\mid A)$
p_B_given_not_A = $P(B\mid \neg A)$

Task:

Compute and return the posterior probability $P(A\mid B)$ .

4) Logistic regression: top-3 features

You are given:

X : a 2D array where each row corresponds to one feature and each column corresponds to one observation (shape: num_features × num_samples )
y : binary outcome array of length num_samples (values in {0,1})
feature_names : array of length num_features

Task:

Fit a logistic regression model to predict y from X (include an intercept).
Rank features by absolute value of their fitted coefficient (exclude the intercept).
Return the names of the top 3 features in descending order of importance.

Notes:

Handle ties deterministically (e.g., break ties by feature name ascending).
Assume inputs are well-formed and numeric.

You are completing a CodeSignal-style assessment (Python or R). Implement solutions for the following four independent questions.

1) Two-sample z-test: required sample size

You are given:

x : numeric array of historical observations for the metric (use it to estimate the metric standard deviation sigma )
alpha : significance level (e.g., 0.05)
power : desired power (e.g., 0.8)
effect_size : the minimum detectable absolute difference in means, $\Delta$

Assumptions:

Two-sided two-sample z-test for a difference in means.
Treatment and control have equal sample size $n$ .
Use $\hat\sigma = \text{std}(x)$ as the population standard deviation estimate.

Task:

Return the minimum integer per-group sample size n required to detect effect_size at level alpha with power .

2) Difference-in-Differences (DiD) + parallel-trend validation

You are given three equal-length arrays:

period[i] : time indicator (contains at least a “pre” and a “post” period; may contain multiple pre periods)
group[i] : 0 = control, 1 = treatment
outcome[i] : numeric outcome

And a numeric threshold for trend validation.

Definitions:

Let $\bar{Y}_{g,t}$ be the mean outcome for group $g\in\{0,1\}$ in period $t$ .
The DiD estimate is:

$\text{DiD} = (\bar{Y}_{1,post}-\bar{Y}_{1,pre}) - (\bar{Y}_{0,post}-\bar{Y}_{0,pre}).$

Parallel-trend / trend validation requirement:

If there are multiple pre periods , compute the group difference $d_t = \bar{Y}_{1,t} - \bar{Y}_{0,t}$ for each pre period $t$ , sort pre periods by time, and validate:

$\max_t |d_{t} - d_{t-1}| \le \text{threshold}.$

If there is only a single pre period, treat trend validation as passing.

Task:

Return (a) the DiD estimate and (b) whether the pre-trend validation passes under the threshold .

3) Bayes’ rule posterior probability

You are given probabilities (as floats) describing an event $A$ and evidence $B$ , such as:

p_A = $P(A)$
p_B_given_A = $P(B\mid A)$
p_B_given_not_A = $P(B\mid \neg A)$

Task:

Compute and return the posterior probability $P(A\mid B)$ .

4) Logistic regression: top-3 features

You are given:

X : a 2D array where each row corresponds to one feature and each column corresponds to one observation (shape: num_features × num_samples )
y : binary outcome array of length num_samples (values in {0,1})
feature_names : array of length num_features

Task:

Fit a logistic regression model to predict y from X (include an intercept).
Rank features by absolute value of their fitted coefficient (exclude the intercept).
Return the names of the top 3 features in descending order of importance.

Notes:

Handle ties deterministically (e.g., break ties by feature name ascending).
Assume inputs are well-formed and numeric.

Implement four DS coding tasks

Quick Overview

1) Two-sample z-test: required sample size

2) Difference-in-Differences (DiD) + parallel-trend validation

3) Bayes’ rule posterior probability

4) Logistic regression: top-3 features

Comments (0)

Implement four DS coding tasks

Quick Overview

1) Two-sample z-test: required sample size

2) Difference-in-Differences (DiD) + parallel-trend validation

3) Bayes’ rule posterior probability

4) Logistic regression: top-3 features

Comments (0)