Analysis

  • Source:OECD Organisation for Economic Co-operation and Development

世界各国の自殺率と各種経済社会指標との重回帰分析 日本の自殺率と失業率の時系列推移

  • 世界各国の自殺率と各種経済社会指標との重回帰分析
  1. 多重共線性は確認していない。
  2. 対象年:2016
MODEL INFO:
Observations: 18
Dependent Variable: SUICIDE:TOT:100000PER
Type: Generalized linear model
  Family: Gamma 
  Link function: log 

MODEL FIT:
χ2(9) = 1.94, p = 0.00
Pseudo-R2 (Cragg-Uhler) = 0.76
Pseudo-R2 (McFadden) = 0.25
AIC = 98.74, BIC = 108.54 

Standard errors: MLE
---------------------------------------------------------------------------------
                                             Est.    2.5%   97.5%   t val.      p
----------------------------------------- ------- ------- ------- -------- ------
(Intercept)                                  4.35   -3.70   12.41     1.06   0.32
UNEMP:TOT:PC_LF                             -0.01   -0.06    0.05    -0.23   0.83
POVERTY:TOT:RT                              -1.16   -5.93    3.61    -0.48   0.65
GDPHRWKD:TOT:USD                             0.00   -0.01    0.02     0.56   0.59
HOUSECOST:PRICERENT:IDX2015                 -0.06   -0.12   -0.00    -2.00   0.08
SHPRICE:TOT:IDX2015                          0.01   -0.00    0.03     1.48   0.18
HOSPITALDISCHARGE:TOT:100000HAB              0.00   -0.00    0.00     1.62   0.14
CHILDVACCIN:MEASLES:PC_CHILD                 0.03   -0.07    0.13     0.63   0.55
FERTILITY:TOT:CHD_WOMAN                     -0.31   -0.75    0.13    -1.38   0.20
POLLUTIONEXP:EXPOS2PM25:MICGRCUBM            0.01   -0.04    0.06     0.43   0.68
---------------------------------------------------------------------------------

Estimated dispersion parameter = 0.08
  1. 個別系列毎のGLM。
  2. 外れ値は除外している。

  • 日本の自殺率と失業率の時系列推移 

lmresult <- lm(formula = tsdata$`JPN:SUICIDE:TOT:100000PER:Diff(lag=1)` ~ tsdata$`JPN:UNEMP:TOT:PC_LF:Diff(lag=1)`)
$`JPN:UNEMP:TOT:PC_LF:Diff(lag=1)`

    Augmented Dickey-Fuller Test

data:  na.omit(x)
Dickey-Fuller = -3.5763, Lag order = 3, p-value = 0.04318
alternative hypothesis: stationary


$`JPN:SUICIDE:TOT:100000PER:Diff(lag=1)`

    Augmented Dickey-Fuller Test

data:  na.omit(x)
Dickey-Fuller = -3.264, Lag order = 3, p-value = 0.08632
alternative hypothesis: stationary
MODEL INFO:
Observations: 56
Dependent Variable: tsdata$`JPN:SUICIDE:TOT:100000PER:Diff(lag=1)`
Type: OLS linear regression 

MODEL FIT:
F(1,54) = 13.13, p = 0.00
R2 = 0.20
Adj. R2 = 0.18 

Standard errors: OLS
--------------------------------------------------------------------------------------
                                                  Est.    2.5%   97.5%   t val.      p
---------------------------------------------- ------- ------- ------- -------- ------
(Intercept)                                      -0.28   -0.61    0.05    -1.69   0.10
tsdata$`JPN:UNEMP:TOT:PC_LF:Diff(lag=1)           2.04    0.91    3.17     3.62   0.00
--------------------------------------------------------------------------------------

    Durbin-Watson test

data:  summary(lmresult)
DW = 1.9801, p-value = 0.4456
alternative hypothesis: true autocorrelation is greater than 0

    One-sample Kolmogorov-Smirnov test

data:  lmresult$residuals
D = 0.1169, p-value = 0.3978
alternative hypothesis: two-sided

    studentized Breusch-Pagan test

data:  lmresult
BP = 7.5477, df = 1, p-value = 0.006009

    Box-Ljung test

data:  lmresult$residuals
X-squared = 0.00058759, df = 1, p-value = 0.9807