# Introduction

It is important to distinguish between the data distribution (aka population distribution) and the sampling distribution. The distinction is critical when working with the central limit theorem or other concepts like the standard deviation and standard error.

In this post we will go over the above concepts and as well as bootstrapping to estimate the sampling distribution. In particular, we will cover the following:

- Data distribution (aka population distribution)
- Sampling distribution
- Central limit theorem (CLT)
- Standard error and its relation with the standard deviation
- Bootstrapping

# Data Distribution

Much of the statistics deals with inferring from samples drawn from a larger population. Hence, we need to distinguish between the analysis done the original data as opposed to analyzing its samples. First, letâ€™s go over the definition of the data distribution:

Letâ€™s first generate random skewed data that will result in a non-normal (non-Gaussian) data distribution. The reason behind generating non-normal data is to better illustrate the relation between data distribution and the sampling distribution.

So, letâ€™s import the Python plotting packages and generate right-skewed data.

```
# Plotting packages and initial setup
import seaborn as sns
="pastel")
sns.set_theme(palette"white")
sns.set_style(
import matplotlib.pyplot as plt
import matplotlib as mpl
"figure.dpi"] = 150 mpl.rcParams[
```

```
# Generate Right-Skewed data set
from scipy.stats import skewnorm
from sklearn.preprocessing import MinMaxScaler
= 10000
num_data_points = 100
max_value = 15 # Positive values are right-skewed
skewness
= skewnorm.rvs(a = skewness,loc=max_value, size=num_data_points, random_state=1)
skewed_random_data = MinMaxScaler().fit_transform(skewed_random_data.reshape(-1, 1))
skewed_data_scaled
# Plot the data (population) distribution
= plt.subplots(figsize=(10, 6))
fig, ax "Data Distribution", fontsize=24, fontweight="bold")
ax.set_title(
=30, stat="density", kde=True, legend=False, ax=ax) sns.histplot(skewed_data_scaled, bins
```

# Sampling Distribution

In the sampling distribution, you draw samples from the dataset and compute a statistic like the mean. Itâ€™s very important to differentiate between the data distribution and the sampling distribution as most confusion comes from the operation done on either the original dataset or its (re)samples.

The algorithm to obtain the sampling distribution is as follows:

- Draw a sample from the dataset.
- Compute a statistic/metric of the drawn sample in Step 1 and save it.
- Repeat Steps 1 and 2 many times.
- Plot the distribution (histogram) of the computed statistic.

```
import numpy as np
import random
= 50
sample_size = []
sample_mean
1) # Setting the seed for reproducibility of the result
random.seed(for _ in range(2000):
= random.sample(skewed_data_scaled.tolist(), sample_size=50)
sample
sample_mean.append(np.mean(sample))
print(f"Mean: {np.mean(sample_mean)} \n")
# Plot the sampling distribution
= plt.subplots(figsize=(10, 6))
fig, ax "Sampling Distribution", fontsize=24, fontweight="bold")
ax.set_title(
=30, stat="density", kde=True, legend=False) sns.histplot(sample_mean, bins
```

`>>> Mean: 0.23269`

Above sampling distribution is basically the histogram of the mean of each drawn sample (in above, we draw samples of 50 elements over 2000 iterations). The mean of the above sampling distribution is around 0.23, as can be noted from computing the mean of all samples means.

# Central Limit Theorem (CLT)

*In CLT, we analyze the sampling distribution and not a data distribution, an important distinction to be made.* CLT is popular in hypothesis testing and confidence interval analysis, and itâ€™s important to be aware of this concept, even though with the use of bootstrap in data science, this theorem is less talked about or considered in the practice of data science (see Bruce and Bruce 2017). More on bootstrapping is provided later in the post.

# Standard Error (SE)

The **standard error** is a metric to describe *the variability of a statistic in the sampling distribution*. We can compute the standard error as follows:

\text{Standard~Error} = SE = \frac{s}{\sqrt{n}}

where s denotes the standard deviation of the sample values and n denotes the sample size. It can be seen from the formula that *as the sample size increases, the SE decreases*.

We can estimate the standard error using the following approach[1]:

- Draw a new sample from a dataset.
- Compute a statistic/metric (e.g., mean) of the drawn sample in Step 1 and save it.
- Repeat Steps 1 and 2 several times.
- An estimate of the standard error is obtained by computing the standard deviation of the previous stepsâ€™ statistics.

While the above approach can be used to estimate the standard error, we can use bootstrapping instead, which is preferable. I will go over that in the next section.

# Bootstrapping

Bootstrapping is an easy way of estimating the sampling distribution by randomly drawing samples from the population (*with replacement*) and computing each resampleâ€™s statistic. Bootstrapping does not depend on the CLT or other assumptions on the distribution, and it is the standard way of estimating SE[1].

Luckily, we can use `[bootstrap()](https://rasbt.github.io/mlxtend/user_guide/evaluate/bootstrap/)`

functionality from the MLxtend library (*You can read my post on MLxtend library covering other interesting functionalities*). This function also provides the flexibility to pass a custom sample statistic.

```
from mlxtend.evaluate import bootstrap
= bootstrap(
avg, std_err, ci_bounds
skewed_data_scaled,=1000,
num_rounds=np.mean, # A function to compute a sample statistic can be passed here
func=0.95,
ci=123 # Setting the seed for reproducibility of the result
seed
)
print(
f"Mean: {avg.round(5)} \n"
f"Standard Error: +/- {std_err.round(5)} \n"
f"CI95: [{ci_bounds[0].round(5)}, {ci_bounds[1].round(5)}]"
)
```

```
>>> Mean: 0.23293
>>> Standard Error: +/- 0.00144
>>> CI95: [0.23023, 0.23601]
```

# Conclusion

The main takeaway is to differentiate between whatever computation you do on the original dataset or the sample of the dataset. Plotting a histogram of the data will result in data distribution, whereas plotting a sample statistic computed over samples of data will result in a sampling distribution. On a similar note, the standard deviation tells us how the data is spread, whereas the standard error tells us how a sample statistic is spread out.

## References

*Practical Statistics for Data Scientists: 50 Essential Concepts*. Oâ€™Reilly Media.

## Citation

```
@online{alizadeh2021,
author = {Esmaeil Alizadeh},
editor = {},
title = {Data {Distribution} Vs. {Sampling} {Distribution:} {What}
{You} {Need} to {Know}},
date = {2021-01-11},
url = {https://ealizadeh.com/blog/statistics-data-vs-sampling-distribution},
langid = {en}
}
```