舆论事件热度对新冠肺炎新增确诊数量有影响吗？

2020-02-21 本文已影响0人 c0659eed21f5

问题

RQ: 舆论事件热度对新冠肺炎新增确诊数量有影响吗？

数据

我们使用知微事见的数据来理解这一研究问题。知微事见,专注于热点事件、企业危机事件、营销事件的研究与分析。当事件符合下列标准之一时，将被收录至知微事见事件库中：1.在短时间内达到高传播量的事件；2.在长期内都保持一定传播量的事件；3.在网络社交媒体中引起热议的事件。

1.什么是事件影响力指数？事件影响力指数（Event Influence Index，EII）是基于全网的社交媒体和网络媒体数据，用来刻画单一事件在互联网上的传播效果的权威指标。指数的计算数据来自全网的社交媒体和网络媒体数据。事件影响力指数是根据事件在社交媒体（以微博、微信为主）和网络媒体上的传播效果进行加和，加和后的事件影响力再通过归一化运算得到范围在0 ~ 100之间的事件影响力指数。

知微事见

可以看到新冠肺炎依然占据舆论热点。

我们使用的数据来自针对新冠肺炎事件脉络的梳理：http://xgml.zhiweidata.net/?from=floating#/

数据来源

数据读取和可视化

import pylab as plt
import pandas as pd
import seaborn as sns
import json

j = json.load(open('zhiwei_line.json'))
df = pd.DataFrame(j)
df.tail()

读取数据

查看一下数据基本类型：

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 52 entries, 0 to 51
Data columns (total 5 columns):

Column Non-Null Count Dtype

0 time 52 non-null object
1 voice 52 non-null object
2 heat 52 non-null object
3 case 52 non-null object
4 allCase 52 non-null object
dtypes: object(5)
memory usage: 2.2+ KB

结果发现都是object，未能被正确识别。

# 将object转化为数值型的类型
df['heat'] = [float(i) for i in df['heat']]
df['case'] = [int(i) for i in df['case']]

可视化

双y坐标轴可视化

# 双y坐标轴可视化
fig = plt.figure(figsize=(12,6),dpi = 200)
plt.style.use('fivethirtyeight')

ax1=fig.add_subplot(111)
ax1.plot(df['time'],  df['heat'], 'r-s')
ax1.set_ylabel('舆论热度', fontsize = 16)
ax1.tick_params(axis='x', rotation=60)
ax1.legend(('舆论热度',),loc='upper left')

ax2=ax1.twinx()
ax2.plot(df['time'], df['case'], 'g-o')
ax2.set_ylabel('新增确诊', fontsize = 16)
ax2.legend(('新增确诊',),loc='upper right')

plt.show()

格兰杰检验

接下来进行格兰杰因果检验。

The Null hypothesis for grangercausalitytests.

H0: the time series in the second column, x2, does NOT Granger cause the time series in the first column, x1.

Grange causality means that past values of x2 have a statistically significant effect on the current value of x1, taking past values of x1 into account as regressors. We reject the null hypothesis that x2 does not Granger cause x1 if the pvalues are below a desired size of the test.

import statsmodels.api as sm
from statsmodels.tsa.stattools import grangercausalitytests
import numpy as np

data = df[21:][['case','heat' ]].pct_change().dropna()
data.plot();

image.png

gc_res = grangercausalitytests(data,4)

Granger Causality
number of lags (no zero) 1
ssr based F test: F=0.0800 , p=0.7796 , df_denom=26, df_num=1
ssr based chi2 test: chi2=0.0892 , p=0.7652 , df=1
likelihood ratio test: chi2=0.0891 , p=0.7653 , df=1
parameter F test: F=0.0800 , p=0.7796 , df_denom=26, df_num=1

Granger Causality
number of lags (no zero) 2
ssr based F test: F=7.5340 , p=0.0030 , df_denom=23, df_num=2
ssr based chi2 test: chi2=18.3436 , p=0.0001 , df=2
likelihood ratio test: chi2=14.1086 , p=0.0009 , df=2
parameter F test: F=7.5340 , p=0.0030 , df_denom=23, df_num=2

Granger Causality
number of lags (no zero) 3
ssr based F test: F=13.4029 , p=0.0001 , df_denom=20, df_num=3
ssr based chi2 test: chi2=54.2818 , p=0.0000 , df=3
likelihood ratio test: chi2=29.7563 , p=0.0000 , df=3
parameter F test: F=13.4029 , p=0.0001 , df_denom=20, df_num=3

Granger Causality
number of lags (no zero) 4
ssr based F test: F=8.8419 , p=0.0005 , df_denom=17, df_num=4
ssr based chi2 test: chi2=54.0915 , p=0.0000 , df=4
likelihood ratio test: chi2=29.2519 , p=0.0000 , df=4
parameter F test: F=8.8419 , p=0.0005 , df_denom=17, df_num=4

结论

舆论热度可以显著地导致新冠肺炎的确诊数量。

练习

请使用格兰杰检验看一下反过来的问题：RQ2: 新冠肺炎确诊数量是否可以导致舆论热度？

思考 🤔

Question: 这是伪相关吗？为什么？不了解伪相关，看这几个例子🌰 http://www.tylervigen.com/spurious-correlations

延伸

Bertram Scheufele, Alexander Haas, Hans-Bernd Brosius （2011）Mirror or Molder? A Study of Media Coverage, Stock Prices, and Trading Volumes in Germany. Journal of Communication, Volume 61, Issue 1, February 2011, Pages 48–70. PDF

This article investigates the short‐term relationship between media coverage, stock prices, and trading volumes of eight listed German companies. A content analysis of news reports about the selected companies and a secondary analysis of the daily changes in closing prices and the trading volumes of these companies were combined in a time‐series design. After ARIMA‐modeling each of them, the results suggest that media coverage rather reflects than shapes the development at stock exchanges from a short‐term perspective (2 months). There were almost no hints for a widespread media effect, that is, an impact on so many investors that it will result in a measurable change in stock prices or trading volumes. Finally, theoretical and methodological consequences for exploring widespread media effects are discussed.