Why statistics quantifies uncertainty in SciPy - Performance Analysis

Time Complexity: Why statistics quantifies uncertainty

O(n)

Understanding Time Complexity

When we use statistics to measure uncertainty, we often run calculations on data samples.

We want to know how the time to do these calculations grows as the data size grows.

Scenario Under Consideration

Analyze the time complexity of the following code snippet.

import numpy as np
from scipy import stats

n = 1000  # Define n before using it

data = np.random.normal(loc=0, scale=1, size=n)
mean = np.mean(data)
conf_int = stats.norm.interval(alpha=0.95, loc=mean, scale=np.std(data)/np.sqrt(n))

This code generates data, calculates the mean, and finds a 95% confidence interval to quantify uncertainty.

Identify Repeating Operations

Identify the loops, recursion, array traversals that repeat.

• Primary operation: Calculating the mean and standard deviation by scanning all data points.
• How many times: Each data point is visited once for mean and once for standard deviation.

How Execution Grows With Input

As the number of data points increases, the time to calculate mean and standard deviation grows proportionally.

Input Size (n)	Approx. Operations
10	About 20 (two passes over 10 points)
100	About 200
1000	About 2000

Pattern observation: The operations grow roughly in direct proportion to the input size.

Final Time Complexity

Time Complexity: O(n)

This means the time to quantify uncertainty grows linearly as the data size grows.

Common Mistake

[X] Wrong: "Calculating uncertainty takes the same time no matter how much data there is."

[OK] Correct: Because mean and standard deviation require looking at every data point, more data means more work.

Interview Connect

Understanding how time grows with data size helps you explain the cost of statistical calculations clearly and confidently.

Self-Check

"What if we used a streaming method that updates mean and variance without storing all data? How would the time complexity change?"

Practice

1. Why do statisticians use confidence intervals when analyzing data?

easy

A. To show a range where the true value likely lies

B. To find the exact value of a measurement

C. To ignore uncertainty in data

D. To make data look more complex

Solution

1. Step 1: Understand the purpose of confidence intervals

   Confidence intervals provide a range of values that likely contain the true population parameter, showing uncertainty.

2. Step 2: Compare options to this purpose

   Only To show a range where the true value likely lies correctly describes this purpose; others misunderstand or ignore uncertainty.

3. Final Answer:

   To show a range where the true value likely lies -> Option A

4. Quick Check:

   Confidence intervals quantify uncertainty = C [OK]

Hint: Confidence intervals give a range, not exact values [OK]

Common Mistakes:

• Thinking confidence intervals give exact values
• Ignoring uncertainty in data
• Confusing confidence intervals with p-values

2. Which of the following is the correct way to import the function for calculating a confidence interval from scipy.stats?

easy

A. from scipy.stats import conf_interval

B. from scipy.stats import t

C. from scipy.stats import tconfint

D. from scipy.stats import t.interval

Solution

1. Step 1: Recall scipy.stats syntax for confidence intervals

   The function to calculate confidence intervals for t-distribution is accessed as t.interval.

2. Step 2: Check import statements

   from scipy.stats import t correctly imports t; others are invalid or incorrect syntax.

3. Final Answer:

   from scipy.stats import t -> Option B

4. Quick Check:

   Correct scipy.stats import syntax = B [OK]

Hint: Use dot notation for submodules like t.interval [OK]

Common Mistakes:

• Trying to import functions that don't exist
• Using wrong function names
• Confusing import syntax with function calls

3. What will be the output of this code snippet?

import numpy as np
from scipy import stats

data = np.array([5, 7, 8, 9, 10])
mean = np.mean(data)
conf_int = stats.t.interval(0.95, len(data)-1, loc=mean, scale=stats.sem(data))
print(conf_int)

medium

A. An error because stats.sem is not defined

B. A single number representing the mean

C. A tuple with two numbers showing the 95% confidence interval

D. A list of all data points

Solution

1. Step 1: Understand the code's purpose

   The code calculates the mean and then the 95% confidence interval using t-distribution and standard error of the mean.

2. Step 2: Analyze the output of stats.t.interval

   This function returns a tuple with lower and upper bounds of the confidence interval, so the print statement outputs a tuple of two numbers.

3. Final Answer:

   A tuple with two numbers showing the 95% confidence interval -> Option C

4. Quick Check:

   stats.t.interval returns confidence bounds = A [OK]

Hint: t.interval returns a tuple (low, high) for confidence [OK]

Common Mistakes:

• Expecting a single number instead of a tuple
• Confusing stats.sem with undefined function
• Thinking output is the raw data list

4. Identify the error in this code that tries to calculate a 90% confidence interval:

from scipy.stats import t

sample = [12, 15, 14, 10, 13]
mean = sum(sample) / len(sample)
interval = t.interval(0.90, len(sample), loc=mean, scale=2)
print(interval)

medium

A. The mean calculation is incorrect

B. The scale parameter cannot be a fixed number like 2

C. The confidence level 0.90 is invalid

D. Degrees of freedom should be len(sample) - 1, not len(sample)

Solution

1. Step 1: Check degrees of freedom parameter

   The degrees of freedom for t-distribution should be sample size minus one, so len(sample) - 1.

2. Step 2: Verify other parameters

   Scale can be a fixed number if it represents standard error; confidence level 0.90 is valid; mean calculation is correct.

3. Final Answer:

   Degrees of freedom should be len(sample) - 1, not len(sample) -> Option D

4. Quick Check:

   Degrees of freedom = sample size - 1 = D [OK]

Hint: Degrees of freedom = sample size minus one [OK]

Common Mistakes:

• Using sample size instead of sample size minus one for df
• Assuming scale must be computed, not fixed
• Misunderstanding confidence level values

5. You have two sets of exam scores from different classes. You want to compare their average scores but also understand how uncertain these averages are. Which approach using scipy best quantifies this uncertainty?

hard

A. Calculate the mean of each class and use stats.t.interval to get confidence intervals for both

B. Calculate the median of each class and compare them directly

C. Use stats.ttest_ind without checking confidence intervals

D. Just compare the highest scores from each class

Solution

1. Step 1: Understand the goal

   You want to compare averages and also quantify uncertainty around those averages.

2. Step 2: Evaluate options for quantifying uncertainty

   Calculating means and confidence intervals with stats.t.interval directly shows uncertainty. Medians or highest scores don't quantify uncertainty. t-test compares means but does not show intervals.

3. Final Answer:

   Calculate the mean of each class and use stats.t.interval to get confidence intervals for both -> Option A

4. Quick Check:

   Confidence intervals quantify uncertainty best = A [OK]

Hint: Confidence intervals show uncertainty around averages [OK]

Common Mistakes:

• Using median without uncertainty measures
• Ignoring confidence intervals in comparisons
• Relying only on highest scores