Table 2 Coefficients and standard errors for linear mixed-effects models for female obesity prevalence

From: Use of social media, search queries, and demographic data to assess obesity prevalence in the United States

 

Twitter variables

Built environment and demographic variables

Twitter and demographic variables

Twitter, demographic and built environment variables

Tweet sentiment toward food

−0.024***(0.005)

 

−0.012***(0.004)

−0.006(0.004)

Percent of tweets food related (2nd tertile)

0.011(0.193)

 

0.101(0.164)

0.267*(0.162)

Percent of tweets food related (3rd tertile)

−1.023***(0.232)

 

−0.693***(0.197)

−0.366*(0.196)

Percent of tweets fast food related (2nd tertile)

0.074(0.176)

 

0.18(0.153)

0.238(0.155)

Percent of tweets fast food related (3rd tertile)

−0.402**(0.194)

 

−0.233(0.173)

−0.131(0.178)

Percent of tweets healthy food related (2nd tertile)

−0.369**(0.185)

 

−0.237(0.158)

−0.106(0.156)

Percent of tweets healthy food related (3rd tertile)

−0.496**(0.210)

 

−0.276(0.179)

−0.118(0.177)

Tweet sentiment toward exercise

−0.010**(0.004)

 

−0.005(0.003)

−0.003(0.003)

Percent of tweets exercise related (2nd tertile)

−1.095***(0.177)

 

−0.386**(0.152)

−0.367**(0.149)

Percent of tweets exercise related (3rd tertile)

−1.782***(0.186)

 

−0.766***(0.163)

−0.679***(0.162)

Median age within female population

 

−0.120***(0.014)

−0.084***(0.014)

−0.101***(0.015)

Percent non-Hispanic black within female population

 

0.136***(0.009)

0.127***(0.006)

0.133***(0.009)

Percent Hispanic within female population

 

−0.006(0.008)

−0.017**(0.007)

−0.006(0.009)

Median household income (in 1000 s)

 

−0.115***(0.009)

−0.131***(-0.006)

−0.115***(-0.01)

Percent with access to exercise space

 

−0.030***(0.003)

 

−0.029***(0.003)

Percent food insecure

 

0.012(0.045)

 

−0.005(0.046)

Percent drive alone to work

 

0.124***(0.011)

 

0.127***(0.012)

Violent crime rate

 

0.0003(0.0004)

 

0.0004(0.0004)

Fast food restaurants (per 1000 residents)

 

−0.110***(0.022)

 

−0.099***(0.023)

Grocery stores (per 1000 residents)

 

−0.045(0.037)

 

−0.07(0.045)

Constant

32.154***(0.615)

31.793***(1.731)

39.792***(0.866)

31.666***(1.795)

Observations

2876

2909

2876

2733

R-Squared

0.564

0.669

0.647

0.669

  1. Note: Coefficients and standard errors (in brackets) for four state-level linear mixed effects models for female obesity prevalence. The models incorporate Twitter variables, demographic variables, and built environment variables. From left to right: (1) a model that includes only Twitter variables, (2) a model that includes demographic and built environment variables, (3) a model that includes Twitter and demographic variables and (4) a model that includes Twitter, demographic and built environment variables. We consistently observed significant negative associations between exercise tweet volume and obesity, and significant negative associations between high food tweet volume and obesity, while controlling for demographic and built environment variables.
  2. *p < 0.1; **p < 0.05; ***p < 0.01
Back to article page