This contains more details on the multiple quantile regression conducted on the health economics data set.
Any questions can be directed to Brendan Lam at blam@hollandbloorview.ca
Variables Included:
- Sex
- Age
- Days Seen from Injury
- Median Income
- Distance from Postal Code
- Number of Symptoms
- Number of Previous Concussions
- Psychiatric History (Y/N)
- Number of Physical Conditions
- Symptom Duration
- Number of Medications
- Number of Therapies Used
Looking at the distribution of each variable with histograms
Missing data was handled with single imputation
## health_econ_data.Psychiatric_hist health_econ_data.DistanceFromPostalCode
## 3 2
## health_econ_data.NumberofSymptoms health_econ_data.NumbPreviousConcuss
## 2 2
## health_econ_data.DaysSeenFromInjury health_econ_data.Duration
## 1 1
Quantile regression for the 0.2th quantile/20th percentile
##
## Call: rq(formula = `Healthcare Cost` ~ ., tau = 0.2, data = qmr_df)
##
## tau: [1] 0.2
##
## Coefficients:
## Value Std. Error t value Pr(>|t|)
## (Intercept) 258.73584 93.26602 2.77417 0.00576
## SexM 16.77248 31.39393 0.53426 0.59343
## `Psychiatric History`Yes -60.62878 43.38568 -1.39744 0.16297
## Age 6.05633 4.54911 1.33132 0.18376
## Income -0.00008 0.00071 -0.11791 0.90619
## Distance 0.03220 0.23377 0.13773 0.89051
## `Number of Symptoms` 36.11734 15.30434 2.35994 0.01870
## `Number of Physical Conditions` 8.27630 20.50314 0.40366 0.68665
## `Number of Meds` 9.53920 11.68967 0.81604 0.41491
## Duration 0.10618 0.10146 1.04656 0.29587
## `Number Other Therapy` 1.88976 14.35727 0.13162 0.89534
Quantile regression for the 0.5th quantile/50th percentile
##
## Call: rq(formula = `Healthcare Cost` ~ ., tau = 0.5, data = qmr_df)
##
## tau: [1] 0.5
##
## Coefficients:
## Value Std. Error t value Pr(>|t|)
## (Intercept) 252.92062 185.97806 1.35995 0.17453
## SexM -77.53343 70.02007 -1.10730 0.26875
## `Psychiatric History`Yes -71.64854 72.40118 -0.98960 0.32290
## Age 13.59783 8.28776 1.64071 0.10156
## Income 0.00045 0.00134 0.33311 0.73921
## Distance 0.26271 0.46821 0.56109 0.57501
## `Number of Symptoms` 81.91633 26.64760 3.07406 0.00224
## `Number of Physical Conditions` 16.57820 54.14332 0.30619 0.75960
## `Number of Meds` 3.38419 15.30344 0.22114 0.82508
## Duration 0.41307 0.10269 4.02249 0.00007
## `Number Other Therapy` 28.81546 18.95451 1.52024 0.12915
Quantile regression for the 0.8th quantile/80th percentile
##
## Call: rq(formula = `Healthcare Cost` ~ ., tau = 0.8, data = qmr_df)
##
## tau: [1] 0.8
##
## Coefficients:
## Value Std. Error t value Pr(>|t|)
## (Intercept) 514.62155 395.67708 1.30061 0.19406
## SexM -145.31391 111.34699 -1.30505 0.19254
## `Psychiatric History`Yes 42.50050 171.46228 0.24787 0.80435
## Age 29.59869 18.61816 1.58978 0.11259
## Income -0.00024 0.00208 -0.11343 0.90974
## Distance 0.14424 0.64486 0.22368 0.82311
## `Number of Symptoms` 147.46042 44.58039 3.30774 0.00102
## `Number of Physical Conditions` 26.41524 64.29961 0.41081 0.68140
## `Number of Meds` -57.62770 21.01749 -2.74189 0.00635
## Duration 0.90221 0.27885 3.23546 0.00130
## `Number Other Therapy` 64.64587 32.85124 1.96784 0.04970
Coefficient contrasts to see if the strength of predictors differs over different quantiles.
## Quantile Regression Analysis of Deviance Table
##
## Model: `Healthcare Cost` ~ Sex + `Psychiatric History` + Age + Income + Distance + `Number of Symptoms` + `Number of Physical Conditions` + `Number of Meds` + Duration + `Number Other Therapy`
## Tests of Equality of Distinct Slopes: tau in { 0.2 0.5 0.8 }
##
## Df Resid Df F value Pr(>F)
## SexM 2 1381 1.5900 0.204293
## `Psychiatric History`Yes 2 1381 0.3113 0.732566
## Age 2 1381 0.9967 0.369363
## Income 2 1381 0.1519 0.859074
## Distance 2 1381 0.1634 0.849237
## `Number of Symptoms` 2 1381 3.6875 0.025281 *
## `Number of Physical Conditions` 2 1381 0.0434 0.957488
## `Number of Meds` 2 1381 5.3811 0.004700 **
## Duration 2 1381 6.0265 0.002478 **
## `Number Other Therapy` 2 1381 2.1379 0.118290
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1