Shift and lag operations in Data Analysis Python - Step-by-Step Execution

Concept Flow - Shift and lag operations

Start with DataFrame

↓

Choose column to shift

↓

Apply shift(n)

↓

Values move down by n rows

↓

Top n rows become NaN

↓

Result: shifted column

↓

Use shifted column for lag analysis

Shift moves data down by n rows, creating lagged versions of columns for time-based comparisons.

Execution Sample

Data Analysis Python

import pandas as pd

df = pd.DataFrame({'value': [10, 20, 30, 40, 50]})
df['lag1'] = df['value'].shift(1)
print(df)

This code creates a lagged column by shifting the 'value' column down by 1 row.

Execution Table

Step	DataFrame 'value' column	Shift(1) applied	Resulting 'lag1' column
Initial	[10, 20, 30, 40, 50]	No	[NaN, NaN, NaN, NaN, NaN]
Apply shift(1)	[10, 20, 30, 40, 50]	Yes	[NaN, 10, 20, 30, 40]
Print df	[10, 20, 30, 40, 50]	Yes	[NaN, 10, 20, 30, 40]

💡 Shift completed: values moved down by 1, top row is NaN because no previous value exists.

Variable Tracker

Variable	Start	After shift(1)	Final
df['value']	[10, 20, 30, 40, 50]	[10, 20, 30, 40, 50]	[10, 20, 30, 40, 50]
df['lag1']	[NaN, NaN, NaN, NaN, NaN]	[NaN, 10, 20, 30, 40]	[NaN, 10, 20, 30, 40]

Key Moments - 2 Insights

Why does the first row of the lagged column become NaN after shift(1)?

Because shift(1) moves values down by one row, the first row has no previous value to fill it, so it becomes NaN as shown in execution_table step 'Apply shift(1)'.

What happens if we use shift(-1) instead of shift(1)?

Shift(-1) moves values up by one row, so the last row becomes NaN because it has no next value. This is the opposite direction of shift(1).

Visual Quiz - 3 Questions

Look at the execution_table at step 'Apply shift(1)'. What is the value of df['lag1'] in the third row?

A 10

B 30

C 20

D NaN

💡 Hint

Check the 'Resulting lag1 column' at step 'Apply shift(1)' in execution_table.

At which step does the 'lag1' column first get its shifted values?

A Initial

B Apply shift(1)

C Print df

D Final

💡 Hint

Look at the 'Shift(1) applied' column in execution_table to see when shift happens.

If we change shift(1) to shift(2), what will happen to the first two rows of 'lag1'?

A They will be NaN

B They will contain the first two values of 'value' column

C They will contain the last two values of 'value' column

D They will be zeros

💡 Hint

Recall that shifting by n moves values down by n rows, so top n rows become NaN.

Concept Snapshot

Shift and lag operations move data up or down in a column.
Use df['col'].shift(n) to move data down by n rows (lag).
Top n rows become NaN because no data exists above.
Useful for comparing current and past values in time series.
Negative n shifts data up (lead).
Always check for NaNs after shifting.

Full Transcript

Shift and lag operations in data science move data in a column up or down by a number of rows. Using pandas in Python, the shift() function moves values down by n rows, creating a lag effect. For example, shift(1) moves all values down by one row, making the first row NaN because it has no previous value. This is useful to compare current data with past data in time series analysis. Negative shifts move data up, creating lead values. After shifting, the new rows created at the edges are filled with NaN to indicate missing data. This visual trace showed step-by-step how the shift function changes the data frame and how variables update.

Practice

1. What does the shift(1) function do to a pandas DataFrame column?

easy

A. Moves the data down by one row, creating a lag of previous values

B. Moves the data up by one row, creating a lead of next values

C. Sorts the column in ascending order

D. Removes the first row of the column

Solution

1. Step 1: Understand the shift function

   The shift(1) moves data down by one row, so each value is replaced by the previous row's value.

2. Step 2: Identify lag vs lead

   Positive shift values create lag (previous rows), so shift(1) creates lag by one row.

3. Final Answer:

   Moves the data down by one row, creating a lag of previous values -> Option A

4. Quick Check:

   shift(1) = lag by 1 row [OK]

Hint: Positive shift means lag (previous rows), negative means lead [OK]

Common Mistakes:

• Confusing shift(1) with shift(-1)
• Thinking shift sorts data
• Assuming shift removes rows

2. Which of the following is the correct syntax to create a new column 'prev_value' that contains the previous row's value of column 'value' in pandas?

easy

A. df['prev_value'] = df['value'].shift(-1)

B. df['prev_value'] = shift(df['value'], 1)

C. df['prev_value'] = df.shift('value', 1)

D. df['prev_value'] = df['value'].shift(1)

Solution

1. Step 1: Recall pandas shift syntax

   The correct syntax is to call shift() on the Series, like df['value'].shift(1).

2. Step 2: Check each option

   df['prev_value'] = df['value'].shift(1) uses correct syntax. df['prev_value'] = df['value'].shift(-1) shifts in the wrong direction. df['prev_value'] = df.shift('value', 1) uses wrong method call. df['prev_value'] = shift(df['value'], 1) uses undefined function.

3. Final Answer:

   df['prev_value'] = df['value'].shift(1) -> Option D

4. Quick Check:

   Correct pandas shift syntax = df['prev_value'] = df['value'].shift(1) [OK]

Hint: Use Series.shift(n) to lag by n rows [OK]

Common Mistakes:

• Using negative shift for previous row
• Calling shift on DataFrame with wrong arguments
• Using undefined shift function

3. Given the DataFrame:

import pandas as pd
df = pd.DataFrame({'value': [10, 20, 30, 40]})
df['lag1'] = df['value'].shift(1)
df['lead1'] = df['value'].shift(-1)
print(df)

What is the output?

medium

A. value lag1 lead1
0 10 10.0 20.0
1 20 20.0 30.0
2 30 30.0 40.0
3 40 40.0 NaN

B. value lag1 lead1
0 10 20.0 NaN
1 20 30.0 10.0
2 30 40.0 20.0
3 40 NaN 30.0

C. value lag1 lead1
0 10 NaN 20.0
1 20 10.0 30.0
2 30 20.0 40.0
3 40 30.0 NaN

D. value lag1 lead1
0 10 NaN NaN
1 20 10.0 20.0
2 30 20.0 30.0
3 40 30.0 40.0

Solution

1. Step 1: Calculate lag1 column

   shift(1) moves values down by one, so first row is NaN, second row gets 10, third 20, fourth 30.

2. Step 2: Calculate lead1 column

   shift(-1) moves values up by one, so last row is NaN, third row gets 40, second 30, first 20.

3. Final Answer:

   value lag1 lead1 0 10 NaN 20.0 1 20 10.0 30.0 2 30 20.0 40.0 3 40 30.0 NaN -> Option C

4. Quick Check:

   Positive shift = lag, negative shift = lead [OK]

Hint: Positive shift adds NaN on top, negative shift adds NaN at bottom [OK]

Common Mistakes:

• Mixing lag and lead directions
• Forgetting NaN appears at shifted edges
• Confusing shift with sorting

4. What is wrong with this code snippet?

df['prev'] = df.shift(1)['value']

Assuming df is a pandas DataFrame with a column 'value'.

medium

A. It works correctly and creates a lagged column

B. It shifts the entire DataFrame, not just the 'value' column, which is inefficient

C. It causes a syntax error due to wrong indexing

D. It will cause a KeyError because shift returns a Series, not DataFrame

Solution

1. Step 1: Understand what df.shift(1) does

   df.shift(1) shifts the entire DataFrame down by one row, including all columns.

2. Step 2: Check indexing and efficiency

   Accessing ['value'] after shift works, but shifting entire DataFrame is inefficient if only one column is needed.

3. Final Answer:

   It shifts the entire DataFrame, not just the 'value' column, which is inefficient -> Option B

4. Quick Check:

   Shift on DataFrame shifts all columns [OK]

Hint: Shift only the needed column, not whole DataFrame [OK]

Common Mistakes:

• Assuming shift returns Series when called on DataFrame
• Expecting syntax error from indexing
• Ignoring performance impact

5. You have a time series DataFrame df with a column 'sales'. You want to create a new column 'pct_change' that shows the percentage change from the previous day. Which code correctly uses shift to calculate this?

hard

A. df['pct_change'] = (df['sales'] - df['sales'].shift(1)) / df['sales'].shift(1) * 100

B. df['pct_change'] = (df['sales'].shift(1) - df['sales']) / df['sales'] * 100

C. df['pct_change'] = df['sales'].pct_change(1)

D. df['pct_change'] = df['sales'].shift(-1) / df['sales'] * 100

Solution

1. Step 1: Understand percentage change formula

   Percentage change = (current - previous) / previous * 100.

2. Step 2: Apply shift to get previous value

   df['sales'].shift(1) gives previous day's sales, so formula in df['pct_change'] = (df['sales'] - df['sales'].shift(1)) / df['sales'].shift(1) * 100 matches the formula.

3. Step 3: Check other options

   df['pct_change'] = (df['sales'].shift(1) - df['sales']) / df['sales'] * 100 reverses numerator and denominator incorrectly. df['pct_change'] = df['sales'].pct_change(1) uses built-in method but question asks for shift usage. df['pct_change'] = df['sales'].shift(-1) / df['sales'] * 100 uses lead instead of lag.

4. Final Answer:

   df['pct_change'] = (df['sales'] - df['sales'].shift(1)) / df['sales'].shift(1) * 100 -> Option A

5. Quick Check:

   Percentage change uses lag with shift(1) [OK]

Hint: Use (current - shift(1)) / shift(1) for percent change [OK]

Common Mistakes:

• Using shift(-1) instead of shift(1)
• Reversing numerator and denominator
• Confusing built-in pct_change with manual shift