Just try to change the definition of PySpark Dataframe:

`import pyspark.sql.functions as F from pyspark.sql.types import StructType, StructField, VariantType

schema = StructType([StructField("statistics", VariantType(), True)])

Initialise simple dataframe with 1 column of raw json data

summary_df = spark.createDataFrame([{'json': '{"key": 123, "data": [4, 5, "str"]}'}])

Create new variant column from raw JSON string

variant_df = summary_df.withColumn("variant", F.parse_json(F.col('json'))) display(variant_df)

print(variant_df.schema)`