High school completion: Moderators of the gap between indiginous and non-indiginous youth
2021-10-22
Summary
Hi folks, Im just cleaning up the file now, making sure that we will be able to directly convert this to a word document
Descriptives (still working on this)
##
## 0 1
## 9781 589##
## 0 1
## 13108 1143##
## 1 2
## 5149 5221##
## 1 2
## 7231 7020## Variable non-Indigenous Indigenous
## 1 Dropout % 15.83% 27.55%
## 2 Cohort 2003 97.96% 2.04%
## 3 Cohort 2009 96.92% 3.08%
## 4 Girls % 50.14% 51.20%
## 5 Urban % 54.16% 45.84%
## 6 Year 10 or Higher % 90.64% 89.98%
## 7 Achievement Index (Mean) -0.09 [-0.13, -0.06] -0.90 [-0.98, -0.82]
## 8 Socioeconomic Status Index (Mean) 0.31 [0.28, 0.33] -0.15 [-0.20, -0.10]use same as paper
Hypothesis set 1:
H1a: Indigenous Australian Children have higher dropout rates
## Joining, by = "term"A tibble: 8 × 5
Variables Log Odds -95% CI +95% CI p
2 Indigenous 0.365 0.129 0.6 0.003 3 Cohort -0.275 -0.413 -0.137 0
4 Grade -0.641 -0.726 -0.556 0
5 Urban -0.478 -0.602 -0.354 0
6 Gender (Boys) 0.509 0.406 0.612 0
7 Attrition Flag 0.613 0.498 0.727 0
8 SES -0.618 -0.686 -0.549 0
0 10 10876 1523 1 4458 1280
H1_gender: Is the gap the same for boys and girls
## Joining, by = "term"## # A tibble: 9 × 5
## Variables `Log Odds` `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <dbl>
## 1 Intercept 4.77 3.88 5.65 0
## 2 Indigenous 0.528 0.154 0.903 0.006
## 3 Gender (Boys) 0.521 0.416 0.626 0
## 4 Cohort -0.275 -0.413 -0.137 0
## 5 Grade -0.642 -0.727 -0.556 0
## 6 Urban -0.478 -0.602 -0.353 0
## 7 Attrition Flag 0.613 0.499 0.727 0
## 8 SES -0.618 -0.686 -0.55 0
## 9 IndXGen -0.313 -0.753 0.126 0.16H1_location: Is the gap the same for urban and rural
## Joining, by = "term"## # A tibble: 9 × 5
## Variables `Log Odds` `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <dbl>
## 1 Intercept 4.78 3.90 5.67 0
## 2 Indigenous 0.19 -0.074 0.453 0.155
## 3 Urban -0.493 -0.619 -0.367 0
## 4 Cohort -0.275 -0.413 -0.137 0
## 5 Grade -0.641 -0.727 -0.556 0
## 6 Gender (Boys) 0.509 0.406 0.612 0
## 7 Attrition Flag 0.613 0.499 0.728 0
## 8 SES -0.618 -0.686 -0.55 0
## 9 IndXLoc 0.413 0.016 0.81 0.042H1_ses: Is the gap the same for Rich and poor and rural
## Rows: 156 Columns: 2## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (2): term, Vars##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.## Joining, by = "term"## # A tibble: 9 × 5
## Variables `Log Odds` `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <dbl>
## 1 Intercept 4.76 3.87 5.64 0
## 2 Indigenous 0.501 0.276 0.726 0
## 3 SES -0.636 -0.706 -0.566 0
## 4 Cohort -0.276 -0.414 -0.138 0
## 5 Grade -0.64 -0.725 -0.554 0
## 6 Urban -0.479 -0.604 -0.354 0
## 7 Gender (Boys) 0.511 0.408 0.614 0
## 8 Attrition Flag 0.614 0.5 0.729 0
## 9 IndXSES 0.584 0.368 0.8 0H1_all three: Gap varies as a function of socioeconomic status, gender and locaion
## Joining, by = "term"## # A tibble: 11 × 5
## Variables `Log Odds` `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <dbl>
## 1 Intercept 4.76 3.87 5.65 0
## 2 Indigenous 0.529 0.166 0.892 0.005
## 3 SES -0.636 -0.706 -0.566 0
## 4 Urban -0.491 -0.617 -0.365 0
## 5 Gender (Boys) 0.524 0.418 0.629 0
## 6 Cohort -0.275 -0.413 -0.137 0
## 7 Grade -0.64 -0.726 -0.555 0
## 8 Attrition Flag 0.616 0.501 0.73 0
## 9 IndXSES 0.553 0.342 0.764 0
## 10 IndXLoc 0.315 -0.065 0.695 0.103
## 11 IndXGen -0.321 -0.707 0.065 0.102hypotheses 1_ cohort interaction only. This gap narrowed as a function on a change in legislation that increased the effective age of first selection to 17 between 2008 and 2010.
## Joining, by = "term"## # A tibble: 9 × 5
## Variables `Log Odds` `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <dbl>
## 1 Intercept 4.76 3.88 5.65 0
## 2 SES -0.618 -0.686 -0.549 0
## 3 Urban -0.479 -0.603 -0.354 0
## 4 Gender (Boys) 0.51 0.406 0.613 0
## 5 Cohort -0.263 -0.406 -0.121 0
## 6 Indigenous 0.532 0.27 0.795 0
## 7 Grade -0.641 -0.726 -0.556 0
## 8 Attrition Flag 0.612 0.497 0.726 0
## 9 IndXCoh -0.317 -0.713 0.079 0.115Hypothesis 1_all three interactions and cohort interaction.
## Joining, by = "term"## # A tibble: 12 × 5
## Variables `Log Odds` `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <dbl>
## 1 Intercept 4.75 3.86 5.64 0
## 2 Indigenous 0.729 0.341 1.12 0
## 3 SES -0.637 -0.707 -0.567 0
## 4 Urban -0.494 -0.62 -0.367 0
## 5 Gender (Boys) 0.524 0.418 0.629 0
## 6 Cohort -0.259 -0.403 -0.115 0.001
## 7 Grade -0.64 -0.725 -0.554 0
## 8 Attrition Flag 0.615 0.5 0.729 0
## 9 IndXSES 0.574 0.365 0.783 0
## 10 IndXLoc 0.358 -0.03 0.745 0.07
## 11 IndXGen -0.321 -0.718 0.076 0.112
## 12 IndXCoh -0.407 -0.791 -0.022 0.039Figures hypothesis 1
hypothesis set 2
Hypothesis 2: Indigenous disadvantage still present when comparing equally advantaged and equally achieving Indigenous and non-Indigenous Youth (controling for achivement)
## # A tibble: 9 × 5
## Parameter Est `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <chr>
## 1 (Intercept) 1.99 1.06 2.92 < .001
## 2 GEO -0.52 -0.648 -0.392 < .001
## 3 ESCS -0.339 -0.411 -0.267 < .001
## 4 GENDER 0.453 0.344 0.563 < .001
## 5 INDIG 0.011 -0.228 0.25 0.926
## 6 COHORT2 -0.339 -0.484 -0.193 < .001
## 7 GRADE -0.378 -0.469 -0.288 < .001
## 8 FLAG_MISS 0.368 0.242 0.494 < .001
## 9 ACH1PV -0.734 -0.802 -0.666 < .001Hypothesis 2_ses. Indiginous gap (controlling for achivement) varies as a function of ses
## # A tibble: 10 × 5
## Parameter Est `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <chr>
## 1 (Intercept) 1.98 1.06 2.91 < .001
## 2 GEO -0.521 -0.649 -0.392 < .001
## 3 GENDER 0.455 0.346 0.564 < .001
## 4 INDIG 0.122 -0.108 0.352 0.292
## 5 ESCS -0.355 -0.429 -0.281 < .001
## 6 COHORT2 -0.339 -0.485 -0.193 < .001
## 7 GRADE -0.378 -0.468 -0.288 < .001
## 8 FLAG_MISS 0.37 0.244 0.496 < .001
## 9 ACH1PV -0.733 -0.801 -0.664 < .001
## 10 INDIG:ESCS 0.478 0.248 0.707 < .001hypothess_2_location
## # A tibble: 10 × 5
## Parameter Est `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <chr>
## 1 (Intercept) 2.00 1.07 2.93 < .001
## 2 GEO -0.538 -0.667 -0.41 < .001
## 3 INDIG -0.196 -0.469 0.077 0.155
## 4 GENDER 0.453 0.344 0.562 < .001
## 5 ESCS -0.339 -0.411 -0.267 < .001
## 6 COHORT2 -0.338 -0.483 -0.193 < .001
## 7 GRADE -0.379 -0.469 -0.289 < .001
## 8 FLAG_MISS 0.369 0.243 0.494 < .001
## 9 ACH1PV -0.735 -0.803 -0.667 < .001
## 10 GEO:INDIG 0.49 0.056 0.923 0.027hypothess_2_gender
## # A tibble: 10 × 5
## Parameter Est `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <chr>
## 1 (Intercept) 1.98 1.06 2.91 < .001
## 2 GEO -0.52 -0.648 -0.391 < .001
## 3 GENDER 0.468 0.357 0.579 < .001
## 4 INDIG 0.209 -0.186 0.604 0.293
## 5 ESCS -0.34 -0.412 -0.268 < .001
## 6 COHORT2 -0.338 -0.484 -0.193 < .001
## 7 GRADE -0.379 -0.469 -0.289 < .001
## 8 FLAG_MISS 0.369 0.243 0.494 < .001
## 9 ACH1PV -0.735 -0.803 -0.667 < .001
## 10 GENDER:INDIG -0.381 -0.846 0.084 0.105Hypotheis2_legislation–This gap narrowed as a function on a change in legislation that increased the effective age of first selection to 17 between 2008 and 2010.
## # A tibble: 10 × 5
## Parameter Est `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <chr>
## 1 (Intercept) 1.98 1.06 2.91 < .001
## 2 GEO -0.521 -0.649 -0.392 < .001
## 3 GENDER 0.454 0.345 0.563 < .001
## 4 COHORT2 -0.329 -0.479 -0.179 < .001
## 5 INDIG 0.148 -0.12 0.415 0.273
## 6 ESCS -0.339 -0.411 -0.267 < .001
## 7 GRADE -0.378 -0.468 -0.288 < .001
## 8 FLAG_MISS 0.367 0.242 0.493 < .001
## 9 ACH1PV -0.734 -0.802 -0.666 < .001
## 10 COHORT2:INDIG -0.258 -0.678 0.163 0.224h2_achivement moderation. Does this gap vary as a function of academic achievement. That is, is the gap between similarly low achieving Indigenous and non-Indigenous children the same as the gap between similarly high achieving Indigenous and non-Indigenous children.
## # A tibble: 10 × 5
## Parameter Est `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <chr>
## 1 (Intercept) 1.98 1.06 2.91 < .001
## 2 GEO -0.522 -0.65 -0.394 < .001
## 3 GENDER 0.455 0.346 0.564 < .001
## 4 COHORT2 -0.341 -0.487 -0.195 < .001
## 5 INDIG 0.444 0.14 0.747 0.005
## 6 ESCS -0.336 -0.409 -0.264 < .001
## 7 GRADE -0.379 -0.469 -0.289 < .001
## 8 FLAG_MISS 0.369 0.244 0.494 < .001
## 9 ACH1PV -0.748 -0.818 -0.678 < .001
## 10 INDIG:ACH1PV 0.388 0.18 0.596 < .001h2_everything in from hyp 1 and 2:
## # A tibble: 14 × 5
## Parameter Est `-95% CI` `+95% CI` p
## <chr> <dbl> <dbl> <dbl> <chr>
## 1 (Intercept) 1.98 1.05 2.91 < .001
## 2 GEO -0.54 -0.669 -0.411 < .001
## 3 INDIG 0.602 0.154 1.05 0.009
## 4 GENDER 0.469 0.358 0.581 < .001
## 5 COHORT2 -0.326 -0.477 -0.175 < .001
## 6 ESCS -0.351 -0.426 -0.277 < .001
## 7 GRADE -0.379 -0.469 -0.288 < .001
## 8 FLAG_MISS 0.371 0.245 0.496 < .001
## 9 ACH1PV -0.745 -0.815 -0.674 < .001
## 10 GEO:INDIG 0.431 0.032 0.83 0.034
## 11 INDIG:GENDER -0.326 -0.73 0.077 0.11
## 12 INDIG:COHORT2 -0.329 -0.714 0.057 0.092
## 13 INDIG:ESCS 0.385 0.162 0.607 < .001
## 14 INDIG:ACH1PV 0.306 0.087 0.526 0.007Figures hypothesis 2
#Figures faceted hypothesis 2: equal achievment
#Graph summarizing hyp1 results
############################indiginous by location
h2_location_1 <- svyglm(DROPOUT ~ GEO*INDIG+INDIG+COHORT+GENDER+GRADE+GEO+ACH1PV+ESCS+FLAG_MISS,design = lsay,family = quasibinomial())
geo_h2 <- ggeffects::ggpredict(h2_location_1, terms = c("INDIG","GEO"))
geo_h2_out <- data.frame(prob = geo_h2$predicted, ci.low = geo_h2$conf.low, ci.high = geo_h2$conf.high,
indig = rep(c("non-Indigenous","Indigenous"),each=2), geo = rep(c(" Prov ","Urban"), 2))
geo_plot_2 <- geo_h2_out %>%
ggplot(aes(x=geo, y=prob, ymin=ci.low, ymax=ci.high)) +
geom_pointrange() +
#geom_hline(yintercept=0, lty=2) + # add a dotted line at x=1 after flip
coord_flip() + # flip coordinates (puts labels on y axis)
xlab("Indigenous") + ylab("Probability") +
facet_wrap(~indig) +
theme_stata() + xlab("") + ylab("Probability of not completing high-school") +
ggtitle("Geography by Indigenous Status")
#geo_plot
##############################indig by cohort
h2_cohort_1 <- svyglm(DROPOUT ~ COHORT*INDIG+INDIG+COHORT+GENDER+GRADE+GEO+ACH1PV+ESCS+FLAG_MISS,design = lsay,family = quasibinomial())
Cohort_h2_1<- ggeffects::ggpredict(h2_cohort_1, terms = c("INDIG","COHORT"))
Cohort_h2_1_out <- data.frame(prob = Cohort_h2_1$predicted, ci.low = Cohort_h2_1$conf.low, ci.high = Cohort_h2_1$conf.high,
indig = rep(c("non-Indigenous","Indigenous"),each=2), Cohort = rep(c(" 2008","2010"), 2))
cohort_plot_2 <- Cohort_h2_1_out %>%
ggplot(aes(x=Cohort, y=prob, ymin=ci.low, ymax=ci.high)) +
geom_pointrange() +
#geom_hline(yintercept=0, lty=2) + # add a dotted line at x=1 after flip
coord_flip() + # flip coordinates (puts labels on y axis)
xlab("Indigenous") + ylab("Probability") +
facet_wrap(~indig) +
theme_stata() + xlab("") + ylab("Probability of not completing high-school") +
ggtitle("Cohort by Indigenous Status")
#two columns one with seperate , one with all
#####ses by indig
h2_SES_1 <- svyglm(DROPOUT ~ ESCS*INDIG+INDIG+COHORT+GENDER+GRADE+GEO+ACH1PV+ESCS+FLAG_MISS,design = lsay,family = quasibinomial())
ses_h2 <- ggeffects::ggpredict(h2_SES_1, terms = c("INDIG","ESCS [-2,-1,0,1,2]"))
ses_h1_out_1 <- data.frame(prob = ses_h2$predicted, ci.low = ses_h2$conf.low, ci.high = ses_h2$conf.high,
indig = rep(c("non-Indigenous","Indigenous"),each=5), ses = rep(-2:2, 2))
ses_plot_2 <- ses_h1_out_1 %>%
ggplot(aes(x=ses, y=prob, ymin=ci.low, ymax=ci.high)) +
geom_pointrange() +
#geom_hline(yintercept=0, lty=2) + # add a dotted line at x=1 after flip
coord_flip() + # flip coordinates (puts labels on y axis)
xlab("Indigenous") + ylab("Probability") +
facet_wrap(~indig) +
theme_stata() + xlab("SES") + ylab("Probability of not completing high-school") +
ggtitle("SES by Indiginous Status")
####Indiginous x achieve
h2_acheive_1 <- svyglm(DROPOUT ~ ACH1PV*INDIG+INDIG+COHORT+GENDER+GRADE+GEO+ACH1PV+ESCS+FLAG_MISS,design = lsay,family = quasibinomial())
Achieve_h2_1<- ggeffects::ggpredict(h2_acheive_1, terms = c("INDIG","ACH1PV [-2,-1,0,1,2]"))
Achieve_h2_out_1 <- data.frame(prob = Achieve_h2_1$predicted, ci.low = Achieve_h2_1$conf.low, ci.high = Achieve_h2_1$conf.high,
indig = rep(c("non-Indigenous","Indigenous"),each=5), ACH1PV = rep(-2:2, 2))
acheive_plot_2 <- Achieve_h2_out_1 %>%
ggplot(aes(x=ACH1PV, y=prob, ymin=ci.low, ymax=ci.high)) +
geom_pointrange() +
#geom_hline(yintercept=0, lty=2) + # add a dotted line at x=1 after flip
coord_flip() + # flip coordinates (puts labels on y axis)
xlab("Indigenous") + ylab("Probability") +
facet_wrap(~indig) +
theme_stata() + xlab("Achievement") + ylab("Probability of not completing high-school") +
ggtitle("Achievement by Indiginous Status")
###face_wrap cohort and indiginous group indigionous face row n =1 line up with each
tmp<- { cohort_plot_2+ geo_plot_2 +ses_plot_2 + acheive_plot_2 + plot_layout(ncol=2)} + plot_layout(ncol=2)
tmp
ggsave("/Users/jociarrochi/Dropbox/to do/lsay/r syntax/hyp2Facet.png",width=10, height=8, dpi=600, tmp)