1.
Preprocessing data
● Outlier: - Replace outliers by median
→ This variable has 3 outliers
- Influential points
- Missing variables
2. Check multicolinearity, correlation matrix
- Multicolinearity - Correlation matrix
3. Linear regression
→ ypred= -0.0845653 sleep_hours_per_night - 0.7284521
mental_health_score + 0.2218116
Avg_daily_usage_hours_clean - 0.7518827 High School +
0.0010471 Undergrduate + 0.9112299 Yes
4. Check for t-test, F-test, Goodness of fit
- F-Test (Overall Significance of the Model)
F(6, 698) = 1394.91, p-value = 0.0000:
The F-test tests the null hypothesis that all regression coefficients (excluding the intercept) are equal to
zero.
Since the p-value is less than 0.05, we reject the null hypothesis.
This indicates that at least one independent variable significantly predicts the dependent variable
(addicted_score), and the model overall is statistically significant.
- T-test
- Goodness of fit
Variables t P>|t| Interpre
tation R-squared = 0.9230: This means that 92.30% of the
variance in the dependent variable addicted_score is
sleep_hours_ - 0.001 Significa explained by the independent variables in the model.
per_night 3. nt This is a very high value, indicating strong explanatory
49 power.
Adjusted R-squared = 0.9224: Also very high, which
mental_healt - 0.000 Significa
23 nt confirms that the model is not overly overfitted despite
h_score
.2 the number of predictors. This value adjusts for the number
3 of variables and observations.
Avg_daily_u 8. 0.000 Significa
sage_hours_c 77 nt
lean
High School - 0.000 Significant
8.28
Undergrduate 0.03 0.975 Not
statistically
significant
Yes 15.7 0.000 Significant
8
5. Check 4 assumptions
[Link] of zero [Link]
- The mean of the residuals is 6.4369, which is significantly different from zero, as shown by the:
p-value < 0.0001 ⇒ strongly reject the null hypothesis.
This means the residuals have a non-zero mean, which violates one of the assumptions in linear regression
(that residuals should have a mean of zero).
- This test evaluates whether the variance of the residuals is constant (i.e., checks for homoskedasticity, an
assumption of OLS regression).
● Null Hypothesis (H₀): Constant variance (homoskedasticity)
● Alternative Hypothesis (H₁): Non-constant variance (heteroskedasticity)
The p-value is much less than 0.05, so you reject the null hypothesis. This indicates strong evidence of
heteroskedasticity, meaning the variance of the residuals is not constant.
- Independent
→ Overall, the scatter diagram of all independent variables has
a trend or repetitive pattern in residuals that would indicate
autocorrelation → violation of the independence assumption.
Normally distributed random variable
Null hypothesis (H₀): Residuals are
Shapiro-Wilk Test normally distributed.
Alternative (H₁): Residuals are not
Since the p-value < 0.05, you reject normally distributed.
the null hypothesis. Residuals do
not follow a normal distribution. Skewness-Kurtosis Test for
Normality
Conclusion: The current regression
model fails to meet foundational ● p-values are < 0.05, the joint
assumptions, so its results and test p-value = 0.0000.
inferences should be interpreted
6. Pratical situations with caution. Therefore, you reject the null
hypothesis → Residuals are not
normally distributed.
Situation 1: A shcool is concerned about the rising issue of social media addiction (or smartphone addiction)
among students. It suspects that poor mental health, lack of sleep, and excessive phone usage are key
contributing factors.
ypred= -0.0845653 sleep_hours_per_night - 0.7284521 mental_health_score + 0.2218116
Avg_daily_usage_hours_clean - 0.7518827 High School + 0.0010471 Undergrduate + 0.9112299 Yes
Mental Health:
A 1-point increase in mental health score lowers addiction score by 0.73 → Poor mental health strongly
increases addiction risk.
Solution: Schools should offer regular mental health screenings, free counseling services, and emotional
regulation workshops to support student well-being and reduce reliance on phones as an emotional escape.
Sleep Hours:
Each extra hour of sleep reduces addiction score by 0.08 → Sleep deprivation plays a role in addiction.
Solution: Schools can include sleep education in wellness programs and promote activities like sports or
meditation to improve sleep and reduce phone dependence.
Daily Phone Use:
Each additional hour of phone use increases addiction score by 0.22 → More screen time = higher addiction.
Solution: Schools can create “Digital Detox” campaigns, use screen trackers for awareness, and expand offline
extracurricular options to help students build healthier tech habits.
Situation 2: A group of parents with children in middle and high school are concerned because their children:
Spend hours glued to their phones (TikTok, YouTube, games),
Show declining academic performance,
Rarely communicate with family,
Are irritable, stressed, and sleep-deprived. They want to understand the underlying causes:
Is it addiction, academic stress, emotional issues, or something else?
Regression Model Analysis:
Teens who sleep less, report lower mental health, and use their phones more tend to have higher addiction
scores. This suggests that excessive screen use may be a coping mechanism for stress or emotional
challenges.
Solution:
Instead of punishing, parents should:
- Open up communication (ask how the child feels instead of criticizing phone use),
- Introduce gentle routines (e.g., no screens before bedtime),
- Encourage healthy offline activities (sports, art, group hobbies),
- Seek help from a counselor or psychologist if needed.
Conclusion:
Based on the model, parents are encouraged to shift from reactive discipline to proactive support, helping
their children build healthier habits and emotional balance.
