pandas functions do not work when using pyspark in jupyter under ubuntu in virtual machine

I am studying pyspark. So, here is how I build up the environment:

1. ubuntu in virtual machine
2. downloading spark 2.4.0
3. install pyspark using pip
4. configuring environment path: 
    export SPARK_HOME="/home/feng/Downloads/spark-2.4.0-bin-hadoop2.7/"
    export PATH=$SPARK_HOME/bin:$PATH
    export PYSPARK_DRIVER_PYTHON=jupyter
    export PYSPARK_DRIVER_PYTHON_OPTS='notebook' 

Then I can use pyspark in jupyter. The first lines are here to discover spark:

import findspark
findspark.init()
import pyspark

Theoretically, I am supposed to use pyspark now. But please see the following two examples:

Example 1:

import findspark
findspark.init()
import pyspark
from pyspark.sql import SparkSession
import matplotlib.pyplot as plt
import pandas as pd

data1 = {'PassengerId': {0: 1, 1: 2, 2: 3, 3: 4, 4: 5},
     'Name': {0: 'Owen', 1: 'Florence', 2: 'Laina', 3: 'Lily', 4: 'William'},
     'Sex': {0: 'male', 1: 'female', 2: 'female', 3: 'female', 4: 'male'},
     'Survived': {0: 0, 1: 1, 2: 1, 3: 1, 4: 0}}

data2 = {'PassengerId': {0: 1, 1: 2, 2: 3, 3: 4, 4: 5},
     'Age': {0: 22, 1: 38, 2: 26, 3: 35, 4: 35},
     'Fare': {0: 7.3, 1: 71.3, 2: 7.9, 3: 53.1, 4: 8.0},
     'Pclass': {0: 3, 1: 1, 2: 3, 3: 1, 4: 3}}

df1_pd = pd.DataFrame(data1, columns=data1.keys())
df2_pd = pd.DataFrame(data2, columns=data2.keys())

df1 = spark.createDataFrame(df1_pd)
df2 = spark.createDataFrame(df2_pd)
df1.show()
df1.filter(df1.Survived ==0 ).show()

Its result is:

+-----------+--------+--------+------+
|PassengerId|    Name|Survived|   Sex|
+-----------+--------+--------+------+
|          1|    Owen|       0|  male|
|          2|Florence|       1|female|
|          3|   Laina|       1|female|
|          4|    Lily|       1|female|
|          5| William|       0|  male|
+-----------+--------+--------+------+

+-----------+-------+--------+----+
|PassengerId|   Name|Survived| Sex|
+-----------+-------+--------+----+
|          1|   Owen|       0|male|
|          5|William|       0|male|
+-----------+-------+--------+----+

Example 2:

import findspark
findspark.init()
import pyspark
from pyspark.sql import SparkSession
import matplotlib.pyplot as plt
import pandas as pd

spark=SparkSession.builder.getOrCreate()
df = spark.read.csv("/home/feng/Downloads/spark-2.4.0-bin-hadoop2.7/examples/src/main/resources/people.csv",
                header=True)
df.show()  


df.filter(df.age > 20 ).show()

The result is:

+------------------+
|      name;age;job|
+------------------+
|Jorge;30;Developer|
|  Bob;32;Developer|
+------------------+

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-4-429ca6587743> in <module>()
 12 
 13 
---> 14 df.filter(df.age > 20 ).show()

/home/feng/Downloads/spark-2.4.0-bin-hadoop2.7/python/pyspark/sql/dataframe.pyc in __getattr__(self, name)
   1298         if name not in self.columns:
   1299             raise AttributeError(
-> 1300                 "'%s' object has no attribute '%s'" % 
(self.__class__.__name__, name))
   1301         jc = self._jdf.apply(name)
   1302         return Column(jc)

AttributeError: 'DataFrame' object has no attribute 'age'

you can see that in these two examples, the function show() has different result. The first one is better then the second one. And for the function filter(), the second one has an error but the first one is good.

I think the only difference (I probably am wrong about this) between these two examples is the first one uses a small data frame generated in the code while the second one read the data from a csv file.

So, what I should do to read data and analyze it properly?