Learn how to analyze relationships between numeric columns with Corr and Cov, draw reproducible random samples with Sample, and compose operations fluently with Pipe.

 

Overview

OperationMethodDescription
CorrelationCorr()Pairwise Pearson correlation matrix
CovarianceCov()Pairwise sample covariance matrix
SamplingSample(n, seed...)Random rows without replacement
ChainingPipe(fn)Apply a custom function in a pipeline

 

 

Sample Data

All examples use this DataFrame:

TempSales
20100
25150
30200
35180
40220

 

Setup Code 

package main

import (
    "fmt"
    "log"

    "github.com/apoplexi24/gpandas"
    "github.com/apoplexi24/gpandas/dataframe"
)

func main() {
    gp := gpandas.GoPandas{}

    df, _ := gp.DataFrame(
        []string{"Temp", "Sales"},
        []gpandas.Column{
            {20.0, 25.0, 30.0, 35.0, 40.0},
            {100.0, 150.0, 200.0, 180.0, 220.0},
        },
        map[string]any{"Temp": gpandas.FloatCol{}, "Sales": gpandas.FloatCol{}},
    )

    // Examples follow...
}

 

 

Corr

Computes the pairwise Pearson correlation matrix over the numeric columns. The result is a square DataFrame whose columns and index are the numeric column names.

 

Function Signature 

func (df *DataFrame) Corr() (*DataFrame, error)

Correlations use rows where both columns are non-null (pairwise complete). A pair with fewer than two overlapping values, or with zero variance, yields NaN.

 

Example 

corr, err := df.Corr()
if err != nil {
    log.Fatalf("Corr failed: %v", err)
}
fmt.Println(corr.String())
+--------------------+--------------------+
| Temp               | Sales              |
+--------------------+--------------------+
| 1                  | 0.9101698143253768 |
| 0.9101698143253768 | 0.9999999999999998 |
+--------------------+--------------------+
[2 rows x 2 columns]

The diagonal is 1 (each column correlates perfectly with itself), and the off-diagonal shows a strong positive correlation between Temp and Sales.

 

 

Cov

Computes the pairwise sample covariance matrix (ddof=1), structured like Corr.

 

Function Signature 

func (df *DataFrame) Cov() (*DataFrame, error)

 

Example 

cov, _ := df.Cov()
fmt.Println(cov.String())
+-------+-------+
| Temp  | Sales |
+-------+-------+
| 62.5  | 337.5 |
| 337.5 | 2200  |
+-------+-------+
[2 rows x 2 columns]

The diagonal holds each column’s variance (e.g., Temp variance is 62.5).

Note: To visualize a correlation matrix as a heatmap, pass the result of Corr() to PlotHeatmap. See Plotting Charts.

 

 

Sample

Returns n rows selected at random without replacement, in random order. An optional seed makes the selection deterministic.

 

Function Signature 

func (df *DataFrame) Sample(n int, seed ...int64) (*DataFrame, error)

 

Example 

s, err := df.Sample(3, 7) // deterministic with seed 7
if err != nil {
    log.Fatalf("Sample failed: %v", err)
}
fmt.Println(s.String())
+------+-------+
| Temp | Sales |
+------+-------+
| 30   | 200   |
| 20   | 100   |
| 40   | 220   |
+------+-------+
[3 rows x 2 columns]

Calling Sample(3, 7) again returns the same rows; omit the seed for a new random sample each time. Index labels of the selected rows are preserved.

 

 

Pipe

Applies a custom function to the DataFrame and returns its result, enabling fluent pipelines of reusable steps.

 

Function Signature 

func (df *DataFrame) Pipe(fn func(*DataFrame) (*DataFrame, error)) (*DataFrame, error)

 

Example 

result, err := df.Pipe(func(d *dataframe.DataFrame) (*dataframe.DataFrame, error) {
    return d.Filter("Sales", dataframe.GreaterThan, 150.0).Result()
})
if err != nil {
    log.Fatalf("Pipe failed: %v", err)
}
fmt.Println(result.String())
+------+-------+
| Temp | Sales |
+------+-------+
| 30   | 200   |
| 35   | 180   |
| 40   | 220   |
+------+-------+
[3 rows x 2 columns]

Pipe is equivalent to calling the function directly, but reads naturally when several steps are composed:

result, err := df.
    Pipe(normalize).
    Pipe(addFeatures)

 

 

Analysis Flow 

flowchart LR
    subgraph DF["DataFrame"]
        D[numeric columns]
    end

    subgraph Stats["Statistics"]
        C[Corr / Cov matrix]
    end

    subgraph Subset["Sampling"]
        S[Sample n rows]
    end

    subgraph Chain["Pipeline"]
        P[Pipe custom steps]
    end

    D --> C
    D --> S
    D --> P

    style DF fill:#1e293b,stroke:#3b82f6,stroke-width:2px
    style Stats fill:#1e293b,stroke:#f59e0b,stroke-width:2px
    style Subset fill:#1e293b,stroke:#8b5cf6,stroke-width:2px
    style Chain fill:#1e293b,stroke:#22c55e,stroke-width:2px

 

 

Error Handling

Common Errors

ErrorCauseSolution
“no numeric columns”Corr/Cov on a non-numeric DataFrameEnsure numeric columns exist
“n … must be in range”Sample(n) with n > row countUse n between 0 and the row count
“fn must not be nil”Pipe(nil)Provide a function

 

 

Thread Safety

These operations read under a read lock and return new DataFrames, leaving the original unchanged.

 

 

See Also