Business Analytics Exam #2

What does it mean to iterate?

To go through the values one at a time

When is for used?

When the number of passes is known (one per row, per item, per file)

The rule with for

The collection says when to stop

How does for work?

for <var> in body>: If there’s any values left in the sequence, assign the next value in <sequence> to <var> and execute the body, then loop back again

What is the sequence in a for code?

What you are going through

<var>

A name you invent. Nothing is special about it.

What does the indented block tell Python in loops?

which lines to repeat (four spaces)

range()

Looping a fixed number of times

list(range(5))

starts at 0, stop before 5 (0, 1, 2, 3, 4)

list(range(1,6))

starts at 1, stops before 6 (1, 2, 3, 4, 5)

list(range(0,10,2))

Every second number up to but not including 10 (0, 2, 4, 6, 8)

len(range(1,6))

How many passes occurred (5)

Where do you place print in loops?

Outside of the indent. That way, it only runs once.

What happens if you place print inside of the indent?

It runs every pass

Accumulator

Name that collects results across all passes. You set it up before the loop and change it inside of the loop.

How is if -> elif -> else tested?

Top to bottom

Where does if -> elif -> else stop?

The first line. Everything below is skipped even if also true

In if -> elif -> else loops, you should put what condition first?

The narrowest

True or false: An inside loop asks its question once per value, and only values that pass get acted on

True

With dictionaries and loops, what should you be mindful of?

Differentiating between values and keys (example: for k, v in price.items() keeps keys and values together)

What happens if you run for k in price

You get keys only as an answer

Building a list using loops

list = [] outside of the loop. At the end of the loop mention list.append(r), with r being "for r in data"

Enumerate

Hands you the position with value (ex: sales = 46, 54, 92. For value in enumerate(sales): print(value) gives (0, 46) (1, 54) (2, 92))

While

The number of passes is not known - repeats until something is true

Rule with while

You must change the condition inside, or it never ends

How does while work?

while <condition>: if <condition> evaluates to true, execute <body> and then loop back.

What must you have in while loops?

A false scenario (ex: x = 5, while x > 0, x = x - 1)

"How long until we double?" is a question answered with what loop?

While (we don't know the answer!)

Parts of a while loop

Something set up before (balance), a condition (balance < 200000), a line inside that moves the condition towards false

Break

Stops the loop entirely, right there

When are breaks used?

When you found out what you came for and the rest doesn't matter

Continue

Skips the rest of this pass and moves to the next value. The loop keeps running.

Nested Loops

The inner loop runs all the way through, once for every pass of the outer one

IndentationError with Loops

Body not indented

NameError with Loops

Accumulator was never started (ex: total = total + r with no total = 0 before the loop)

IndexError with Loops

Last pass asks for a position that isn't there

ValueError with loops

Doesn't specify .items() in dictionary loops

Business Analytics Process

Question/Problem, Data Collection, Processing/Cleaning, Analytics Methods, Decision

Business Analytics

Process of transforming data into actionable insights by identifying meaningful trends, making predictions, and other techniques

Levels of Qualitative Data

Nominal and Ordinal

Nominal Data

Names (may compute counts, mode, shares)

Ordinal Data

Names with order (may compute counts, mode, median)

Levels of Quantitative Data

Discrete and Continuous

Discrete Data

Whole Numbers

Continuous Data

Numbers with Decimals

Types of Continuous Data

Interval and Ratio

Interval Data

No true zero

Ratio Data

True Zero

Time-Series Data

Data collected over many time periods

Cross-Sectional Data

Data collected from several entities at the same time

Name the skew: mean < median < mode

Negative/Left Skewed

Name the skew: mean = median = mode

Normal Distribution

Name the skew: mode < median < mean

Positive/Right skewed

What method of central tendency should you quote in positive/negative skews?

Median

What method of central tendency should you quote in normal distribution?

Mean

Methods of dispersion

Range, variance, standard deviation

The larger the spread, the (larger/smaller) the dispersion

larger

What happens when standard deviation = 0?

Every value is identical

df.info()

Gives non-null counts, data types, shape

df.describe()

Gives count, mean, standard deviation, minimum, quantiles, maximum (numeric columns only)

df["amount"].quantile(0.25)

A quarter of the data is below this number

df["amount"].quantile(0.75)

A quarter of the data is above this number

Which descriptive statistics are not resistance to "bad rows"/outliers?

Mean, standard deviation

Which descriptive statistics are resistant to "bad rows"/outliers?

Median, IQR

df["names"].value_counts()

How many of each value there are

df.groupby("names")["amount"].mean()

Mean of amount per name in names

df.kurt()

How much lives out in the tails

df["names"].nunique()

How many distinct values there are

df.iloc[3]

Gives 3rd index row information

Difference between iloc and loc

iloc doesn't survive a sort while loc does

df.iloc[-1]

Gives last row information

df.iloc[3,2]

Gives one cell by two positions (can also use position, column name)

df.sort_values("value")

Sorts by smallest first (ascending=False reverses this)

df.sort_index()

Back to index order

Primary sources

First-hand gathered data (interviews, surveys, experiments, observations)

Secondary Sources

Collected by others (government data sources, journals, public websites)

4 Dimensions of Data Quality

Accuracy, Completeness, Consistency, Usability

Questions to ask about data

Who collected/produced it? When was the data collected? What was their intent? is it a reputable source?

What happens if you have biased data?

Get new data

What happens if you have outliers?

Exclude, include, or investigate

What happens if you have invalid data?

Fix or remove

What happens if you have missing data?

Drop rows/columns or replace (with mean, median)

What happens if you have duplicates?

Remove by .drop_duplicates()

What happens if you have stale data?

Be cautious with results

What happens if you have inconsistency?

Remove or fix by str.strip(), str.title(), converting dates

Data Cleaning Process Steps

Copy original data, look using info, strip symbols, convert types, drop duplicates, handle blanks, look again

df.shape()

Gives rows and columns as a tuple

df.isna().sum()

How many blanks are there?

df.duplicated().sum()

How many duplicates are there?

df['names'].nunique()

How many distinct spellings are there?

list(df['names'].unique())

What are the distinct spellings in the column 'names'?

copy = df.copy()

Creates a copy of the original dataset

df["column"].astype(float/int/str)

Fixes data types

df['pop'] = df['pop'].fillna("value")

Fills values with "value"

df = df.dropna(axis = 0, how = 'all')

Drops a row (to drop a column, change axis to 1)

df[df.duplicated()]

Point out duplicated rows

df["amount"] = pd.to_numeric(df["amount"], errors ="coerce")

Use if there's data issues while you're trying to convert data types (ex: there's a variable named "pending amongst a sea of numbers). Makes the variable nan.

iqr = df["pop"].quantile(0.75) - df["pop].quantile(0.25)

Finding IQR

max = df["pop"].quantile(0.25) - iqr and min = df["pop"].quantile(0.75) + iqr

Finding the maximum and minimum values

(df[df["pop"] < min]) | (df["pop"] > max)])

Finding outliers