Enso
- SECTION I: CHOOSING THE BEST ITEMS FOR EACH OCCASION
- PART 1: Choosing The Best Reflection Survey Items For Each Occasion
- Which Perception Items (Self v. Unit v. UVA) predict Wellness Screen scales at each occasion?
- SELF Perception predicts PCL2-Sum (at each occasion)
- UNIT Perception predicts AUDIT3-Sum (at each occasion)
- No perception items predict PHQ2-Sum (at each occasion)
- No perception items predict GAD2-sum (at each occasion)
- NO analysis for Maslauch9 composite and subcomposites as I have no information on how to score.
- Which Perception Items (Self vs. Unit vs. UVA) predict full PQL and PSS Scale at each occasion?
- Which Perception Items (Self v. Unit v. UVA) predict Wellness Screen scales at each occasion?
- Part 2: Choosing the Best Wellness Survey Items
- PART 1: Choosing The Best Reflection Survey Items For Each Occasion
- SECTION II: CHOOSING THE BEST ITEMS FOR EACH PERSON ACROSS TIME
- Part 1: Choosing The Best Reflection Survey Items For Each Person Over Time
- Which Perception Items (Self v. Unit v. UVA) predict Wellness Screen scales for each person?
- SELF Perception predicts PCL2-Sum (for each person)
- UNIT Perception predicts AUDIT3-Sum (for each person)
- No perception items predict PHQ2-Sum (for each person)
- No perception items predict GAD2-sum (for each person)
- NO analysis for Maslauch9 composite and subcomposites as I have no information on how to score.
- Which Perception Items (Self v. Unit v. UVA) predict full PQL and PSS scales for each person?
- UNIT Perception predicts Maslauch COMPASSION (for each participant across time)
- UNIT Perception predicts Maslauch BURNOUT (for each participant across time)
- UNIT Perception predicts PSS Stress (For each participant across time)
- No Perception Items predict PQL Stress (For each participant across time)
- Which Perception Items (Self v. Unit v. UVA) predict Wellness Screen scales for each person?
- Part 2: Choosing The Best Wellness Screeen Items For Each Person Over Time
- Which Wellness Survey items predict full PQL and PSS scales for each person across time.
- GAD2-Sum predicts PSS Stress (For each participant across time)
- Maslauch-9(Item3) predicts PQL-Stress (For each participant across time)
- Maslauch-9(Items 4&9) predict Maslauch COMPASSION (For each participant across time)
- GAD2-Sum, PCL2-Sum, Maslauch-9(Item4) predict Maslauch BURNOUT (For each participant across time)
- Which Wellness Survey items predict full PQL and PSS scales for each person across time.
- Part 1: Choosing The Best Reflection Survey Items For Each Person Over Time
- SECTION III: CHOOSING THE BEST SCALES/ITEMS THAT PREDICT PERCEPTION (SELF, UNIT, & UVA) ACROSS TIME
The analyses presented here use an advanced machine learning technique (MARS) in order to identify the most predictive variables in a set.
The analyses include three items from the Reflection Survey (Self, Unit, and UVA), all items from the Wellness Screen, and the composites of PQL-Burnout, PQL-Compassion, PQL-Stress, and PSS-Stress as full measures. Predictors are always analyzed in a set with the other items administered in that survey (i.e., Reflection Survey, Wellness Screen, or the survey with the complete item set for PQL and PSS).
The Reflection Survey items are always analyzed individually. The Wellness Screen items are analyzed as single items when used as predictors (for the purpose of item reduction) and as composites when used as outcomes (for the purpose of best measurement of the underlying constructs). The PQL and PSS are always analyzed as composites (as they are validated scales and item reduction is not of interest).
All of these analyses should be interpreted with caution due to the small sample (29 people and less than 50 occasions). Variables/surveys not included contained too few responses for meaningful analyses. Only three items from the Reflection Survey (Self, Unit, and UVA) are included as the other items had too few data points.
The analyses are separated into three sections.
In Section I, we explore the question of which single items as answered at a particular time can be used to decide whether or not to administer additional items at that same time. This is the key analysis if you are interested in selecting fewer items from the Reflection or Wellness Surveys each time you administer a survey. No clear winners from the Reflection Survey and the questions on the brief Wellness survey generally predict the corresponding scales on the PQL and PSS scales.
In Section II, we explore a similar question but by averaging scores across time and looking at which items best predict full scales across all surveys taken by a participant. This tells you whether a participant’s average response on a single item over time is a good predictor of their average response on constructs such as stress and burnout. You should see that as an additional analysis to Section I. Unit (vs. UVA and Self) perception is a clear winner in those analyses.
In Section III, we ask which full scales predict participants’ perception of Self, Unit, and UVA across time. This can tell you what full scales to keep if you are interested in perception of Self, Unit, and UVA as the key outcomes. These analyses are only presented at the person level (averaged across time) as this seems to be the key question here.
For effect size estimates, you can look at the RSq, which stands for R-squared, which can be interpreted as the variance explained in the outcome by the predictor(s) that were selected. In each analysis, the RSq is under ‘coefficients’ in the last row, right before the call to ‘plotmo(mars)’. You can also look at GRSq (Generalized R Squared), which will always be smaller than the RSq. The GRSq is a measure of how likely you are to find the same pattern of findings in a new sample. The difference between GRSq and RSq can be thought of as the uncertainty present in the current findings, which is substantial because the sample is not very large. For models where 2 or more predictors were selected, RSq and GRSq will represent the variance explained by all of the predictors taken together, so if you want an effect size for each of them, additional models will need to be run.
SECTION I: CHOOSING THE BEST ITEMS FOR EACH OCCASION
PART 1: Choosing The Best Reflection Survey Items For Each Occasion
Which Perception Items (Self v. Unit v. UVA) predict Wellness Screen scales at each occasion?
SELF Perception predicts PCL2-Sum (at each occasion)
mars = earth(Wellness.PCL.Sum ~ Refl.Self + Refl.Unit + Refl.UVA, data = dl,
pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=Wellness.PCL.Sum~Refl.Self+Refl.Unit+Refl.UVA,
data=dl, pmethod=“backward”)
coefficients(Intercept) 0.967
h(Refl.Self-1.5) 1.114
Selected 2 of 12 terms, and 1 of 3 predictors
Termination condition: RSq changed by less than 0.001 at 12 terms
Importance: Refl.Self, Refl.Unit-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 3.16 RSS 124 GRSq 0.0133 RSq 0.101
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.1 1.5 1.5
UNIT Perception predicts AUDIT3-Sum (at each occasion)
mars = earth(Wellness.AUDIT.Sum ~ Refl.Self + Refl.Unit + Refl.UVA, data = dl,
pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=Wellness.AUDIT.Sum~Refl.Self+Refl.Unit+Refl.UVA,
data=dl, pmethod=“backward”)
coefficients(Intercept) 2.038
h(Refl.Unit-1) 0.975
Selected 2 of 14 terms, and 1 of 3 predictors
Termination condition: Reached nk 21
Importance: Refl.Unit, Refl.Self-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 4.64 RSS 182 GRSq 0.0122 RSq 0.1
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.1 1.5 1.5
No perception items predict PHQ2-Sum (at each occasion)
mars = earth(Wellness.PHQ.Sum ~ Refl.Self + Refl.Unit + Refl.UVA, data = dl,
pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=Wellness.PHQ.Sum~Refl.Self+Refl.Unit+Refl.UVA,
data=dl, pmethod=“backward”)
coefficients(Intercept) 1.16
Selected 1 of 14 terms, and 0 of 3 predictors
Termination condition: Reached nk 21
Importance: Refl.Self-unused, Refl.Unit-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 (intercept only model)
GCV 2.42 RSS 104 GRSq 0 RSq 0
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.1 1.5 1.5
No perception items predict GAD2-sum (at each occasion)
mars = earth(Wellness.GAD.Sum ~ Refl.Self + Refl.Unit + Refl.UVA, data = dl,
pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=Wellness.GAD.Sum~Refl.Self+Refl.Unit+Refl.UVA,
data=dl, pmethod=“backward”)
coefficients(Intercept) 1.84
Selected 1 of 14 terms, and 0 of 3 predictors
Termination condition: Reached nk 21
Importance: Refl.Self-unused, Refl.Unit-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 (intercept only model)
GCV 3.21 RSS 138 GRSq 0 RSq 0
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.1 1.5 1.5
NO analysis for Maslauch9 composite and subcomposites as I have no information on how to score.
Which Perception Items (Self vs. Unit vs. UVA) predict full PQL and PSS Scale at each occasion?
UVA Perception predicts Maslauch COMPASSION (at each occasion)
mars = earth(PQL.Compassion ~ Refl.Self + Refl.Unit + Refl.UVA, data = dll,
pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=PQL.Compassion~Refl.Self+Refl.Unit+Refl.UVA, data=dll,
pmethod=“backward”)
coefficients(Intercept) 37.91
h(1-Refl.UVA) 7.81
Selected 2 of 14 terms, and 1 of 3 predictors
Termination condition: Reached nk 21
Importance: Refl.UVA, Refl.Self-unused, Refl.Unit-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 41.6 RSS 1299 GRSq 0.0466 RSq 0.15
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.1 1.5 1.1
UVA Perception predicts PSS Stress (at each occasion)
mars = earth(PSS.Stress ~ Refl.Self + Refl.Unit + Refl.UVA, data = dll, pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=PSS.Stress~Refl.Self+Refl.Unit+Refl.UVA, data=dll,
pmethod=“backward”)
coefficients(Intercept) 14.10
h(Refl.UVA-1.7) 7.14
Selected 2 of 14 terms, and 1 of 3 predictors
Termination condition: Reached nk 21
Importance: Refl.UVA, Refl.Self-unused, Refl.Unit-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 51 RSS 1595 GRSq 0.0321 RSq 0.137
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.1 1.5 1.1
UNIT and UVA Perception predict PQL Stress (at each occasion)
mars = earth(PQL.Stress ~ Refl.Self + Refl.Unit + Refl.UVA, data = dll, pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=PQL.Stress~Refl.Self+Refl.Unit+Refl.UVA, data=dll,
pmethod=“backward”)
coefficients(Intercept) 18.75
h(Refl.Unit-1.6) 24.00
h(Refl.Unit-2) -27.23
h(Refl.UVA-1.4) -8.42
Selected 4 of 14 terms, and 2 of 3 predictors
Termination condition: Reached nk 21
Importance: Refl.UVA, Refl.Unit, Refl.Self-unused
Number of terms at each degree of interaction: 1 3 (additive model)
GCV 35.2 RSS 856 GRSq 0.123 RSq 0.391
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.1 1.5 1.1
Part 2: Choosing the Best Wellness Survey Items
Which Wellness Survey items predict full PQL and PSS scales at each occasion?
Maslauch-9(Items 4 & 9) and AUDIT (Item 3) predict PQL-COMPASSION (at each occasion)
mars = earth(PQL.Compassion ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dll, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=PQL.Compassion~Wellness.PHQ2.1+Wellness.PHQ2.1+Wel…),
data=dll, pmethod=“backward”)
coefficients(Intercept) 34.52
Wellness.AUDIT.3 -2.40
h(2-Wellness.Masl9.4) 4.07
h(Wellness.Masl9.9-4) 4.27
Selected 4 of 16 terms, and 3 of 17 predictors
Termination condition: GRSq -10 at 16 terms
Importance: Wellness.Masl9.4, Wellness.Masl9.9, …
Number of terms at each degree of interaction: 1 3 (additive model)
GCV 31.1 RSS 756 GRSq 0.287 RSq 0.505
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 1 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0 0 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
GAD (Item 2), PCL (Item 1), and Maslauch9 (Items 4 & 9) predict PQL-BURNOUT (at each occasion)
mars = earth(PQL.Burnout ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dll, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=PQL.Burnout~Wellness.PHQ2.1+Wellness.PHQ2.1+Wellne…),
data=dll, pmethod=“backward”)
coefficients(Intercept) 25.25
Wellness.PCL2.1 1.31
h(1-Wellness.GAD2.2) -4.41
h(3-Wellness.Masl9.4) -2.49
h(4-Wellness.Masl9.9) 1.84
Selected 5 of 15 terms, and 4 of 17 predictors
Termination condition: GRSq -10 at 15 terms
Importance: Wellness.GAD2.2, Wellness.Masl9.4, Wellness.Masl9.9, …
Number of terms at each degree of interaction: 1 4 (additive model)
GCV 14.3 RSS 304 GRSq 0.535 RSq 0.719
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 1 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0 0 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
Maslauch-9 (Item7) predicts PQL-Stress (at each occasion)
mars = earth(PQL.Stress ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dll, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=PQL.Stress~Wellness.PHQ2.1+Wellness.PHQ2.1+Wellnes…),
data=dll, pmethod=“backward”)
coefficients(Intercept) 13.08
Wellness.AUDIT.3 -5.20
h(Wellness.GAD2.2-1) 3.56
h(Wellness.AUDIT.1-1) 4.64
h(Wellness.Masl9.7-1) 2.35
Selected 5 of 16 terms, and 4 of 17 predictors
Termination condition: GRSq -10 at 16 terms
Importance: Wellness.Masl9.7, Wellness.AUDIT.1, Wellness.AUDIT.3, …
Number of terms at each degree of interaction: 1 4 (additive model)
GCV 24.2 RSS 514 GRSq 0.395 RSq 0.634
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 1 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0 0 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
GAD2 (Item 2) predicts PSS-Stress (at each occasion)
mars = earth(PSS.Stress ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dll, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=PSS.Stress~Wellness.PHQ2.1+Wellness.PHQ2.1+Wellnes…),
data=dll, pmethod=“backward”)
coefficients(Intercept) 16.49
h(1-Wellness.GAD2.2) -6.67
h(Wellness.GAD2.2-1) 5.90
Selected 3 of 17 terms, and 1 of 17 predictors
Termination condition: GRSq -Inf at 17 terms
Importance: Wellness.GAD2.2, Wellness.PHQ2.1-unused, …
Number of terms at each degree of interaction: 1 2 (additive model)
GCV 19.6 RSS 543 GRSq 0.628 RSq 0.706
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 1 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0 0 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
SECTION II: CHOOSING THE BEST ITEMS FOR EACH PERSON ACROSS TIME
Part 1: Choosing The Best Reflection Survey Items For Each Person Over Time
Which Perception Items (Self v. Unit v. UVA) predict Wellness Screen scales for each person?
SELF Perception predicts PCL2-Sum (for each person)
mars = earth(Wellness.PCL.Sum ~ Refl.Self + Refl.Unit + Refl.UVA, data = dw,
pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=Wellness.PCL.Sum~Refl.Self+Refl.Unit+Refl.UVA,
data=dw, pmethod=“backward”)
coefficients(Intercept) 0.89
h(Refl.Unit-1.15385) 1.92
Selected 2 of 9 terms, and 1 of 3 predictors
Termination condition: RSq changed by less than 0.001 at 9 terms
Importance: Refl.Unit, Refl.Self-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 2.67 RSS 59.6 GRSq 0.149 RSq 0.27
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.4 1.2 1.5
UNIT Perception predicts AUDIT3-Sum (for each person)
mars = earth(Wellness.AUDIT.Sum ~ Refl.Self + Refl.Unit + Refl.UVA, data = dw,
pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=Wellness.AUDIT.Sum~Refl.Self+Refl.Unit+Refl.UVA,
data=dw, pmethod=“backward”)
coefficients(Intercept) 2.55
Selected 1 of 11 terms, and 0 of 3 predictors
Termination condition: GRSq -10 at 11 terms
Importance: Refl.Self-unused, Refl.Unit-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 (intercept only model)
GCV 5.83 RSS 152 GRSq 0 RSq 0
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.4 1.2 1.5
No perception items predict PHQ2-Sum (for each person)
mars = earth(Wellness.PHQ.Sum ~ Refl.Self + Refl.Unit + Refl.UVA, data = dw,
pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=Wellness.PHQ.Sum~Refl.Self+Refl.Unit+Refl.UVA,
data=dw, pmethod=“backward”)
coefficients(Intercept) 0.99
h(Refl.UVA-1.6625) 2.74
Selected 2 of 11 terms, and 1 of 3 predictors
Termination condition: GRSq -10 at 11 terms
Importance: Refl.UVA, Refl.Self-unused, Refl.Unit-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 2.55 RSS 56.8 GRSq 0.0974 RSq 0.226
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.4 1.2 1.5
No perception items predict GAD2-sum (for each person)
mars = earth(Wellness.GAD.Sum ~ Refl.Self + Refl.Unit + Refl.UVA, data = dw,
pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=Wellness.GAD.Sum~Refl.Self+Refl.Unit+Refl.UVA,
data=dw, pmethod=“backward”)
coefficients(Intercept) 1.36
h(Refl.UVA-1.45) 2.81
Selected 2 of 11 terms, and 1 of 3 predictors
Termination condition: GRSq -10 at 11 terms
Importance: Refl.UVA, Refl.Self-unused, Refl.Unit-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 3.33 RSS 74.4 GRSq 0.144 RSq 0.266
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.4 1.2 1.5
NO analysis for Maslauch9 composite and subcomposites as I have no information on how to score.
Which Perception Items (Self v. Unit v. UVA) predict full PQL and PSS scales for each person?
UNIT Perception predicts Maslauch COMPASSION (for each participant across time)
mars = earth(PQL.Compassion ~ Refl.Self + Refl.Unit + Refl.UVA, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=PQL.Compassion~Refl.Self+Refl.Unit+Refl.UVA, data=dw,
pmethod=“backward”)
coefficients(Intercept) 37.69
h(1.45-Refl.Unit) 5.46
Selected 2 of 11 terms, and 1 of 3 predictors
Termination condition: GRSq -10 at 11 terms
Importance: Refl.Unit, Refl.Self-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 36.4 RSS 813 GRSq 0.0278 RSq 0.166
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.4 1.2 1.5
UNIT Perception predicts Maslauch BURNOUT (for each participant across time)
mars = earth(PQL.Burnout ~ Refl.Self + Refl.Unit + Refl.UVA, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=PQL.Burnout~Refl.Self+Refl.Unit+Refl.UVA, data=dw,
pmethod=“backward”)
coefficients(Intercept) 24.75
h(1.45-Refl.Unit) -6.07
Selected 2 of 10 terms, and 1 of 3 predictors
Termination condition: RSq changed by less than 0.001 at 10 terms
Importance: Refl.Unit, Refl.Self-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 25.8 RSS 575 GRSq 0.135 RSq 0.258
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.4 1.2 1.5
UNIT Perception predicts PSS Stress (For each participant across time)
mars = earth(PSS.Stress ~ Refl.Self + Refl.Unit + Refl.UVA, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=PSS.Stress~Refl.Self+Refl.Unit+Refl.UVA, data=dw,
pmethod=“backward”)
coefficients(Intercept) 13.38
h(Refl.Unit-1.15385) 6.73
Selected 2 of 11 terms, and 1 of 3 predictors
Termination condition: GRSq -10 at 11 terms
Importance: Refl.Unit, Refl.Self-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 44.9 RSS 1002 GRSq 0.0823 RSq 0.213
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.4 1.2 1.5
No Perception Items predict PQL Stress (For each participant across time)
mars = earth(PQL.Stress ~ Refl.Self + Refl.Unit + Refl.UVA, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call: earth(formula=PQL.Stress~Refl.Self+Refl.Unit+Refl.UVA, data=dw,
pmethod=“backward”)
coefficients(Intercept) 18.8
Selected 1 of 11 terms, and 0 of 3 predictors
Termination condition: GRSq -10 at 11 terms
Importance: Refl.Self-unused, Refl.Unit-unused, Refl.UVA-unused
Number of terms at each degree of interaction: 1 (intercept only model)
GCV 33.7 RSS 877 GRSq 0 RSq 0
plotmo(mars)plotmo grid: Refl.Self Refl.Unit Refl.UVA
1.4 1.2 1.5
Part 2: Choosing The Best Wellness Screeen Items For Each Person Over Time
Which Wellness Survey items predict full PQL and PSS scales for each person across time.
GAD2-Sum predicts PSS Stress (For each participant across time)
mars = earth(PSS.Stress ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=PSS.Stress~Wellness.PHQ2.1+Wellness.PHQ2.1+Wellnes…),
data=dw, pmethod=“backward”)
coefficients(Intercept) 16.79
h(1-Wellness.GAD2.2) -7.80
h(Wellness.GAD2.2-1) 4.84
Selected 3 of 13 terms, and 1 of 17 predictors
Termination condition: GRSq -Inf at 13 terms
Importance: Wellness.GAD2.2, Wellness.PHQ2.1-unused, …
Number of terms at each degree of interaction: 1 2 (additive model)
GCV 14 RSS 264 GRSq 0.714 RSq 0.793
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 0.88 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0.25 1 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5.2 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
Maslauch-9(Item3) predicts PQL-Stress (For each participant across time)
mars = earth(PQL.Stress ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=PQL.Stress~Wellness.PHQ2.1+Wellness.PHQ2.1+Wellnes…),
data=dw, pmethod=“backward”)
coefficients(Intercept) 16.87
h(Wellness.Masl9.3-2) 2.05
Selected 2 of 11 terms, and 1 of 17 predictors
Termination condition: GRSq -10 at 11 terms
Importance: Wellness.Masl9.3, Wellness.PHQ2.1-unused, …
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 29 RSS 648 GRSq 0.138 RSq 0.261
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 0.88 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0.25 1 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5.2 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
Maslauch-9(Items 4&9) predict Maslauch COMPASSION (For each participant across time)
mars = earth(PQL.Compassion ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=PQL.Compassion~Wellness.PHQ2.1+Wellness.PHQ2.1+Wel…),
data=dw, pmethod=“backward”)
coefficients(Intercept) 32.76
h(1-Wellness.PCL2.1) 4.13
h(Wellness.AUDIT.1-2) -4.80
h(4-Wellness.Masl9.9) 4.47
h(Wellness.Masl9.9-4) 6.52
Selected 5 of 12 terms, and 3 of 17 predictors
Termination condition: GRSq -10 at 12 terms
Importance: Wellness.Masl9.9, Wellness.AUDIT.1, Wellness.PCL2.1, …
Number of terms at each degree of interaction: 1 4 (additive model)
GCV 32.1 RSS 414 GRSq 0.144 RSq 0.576
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 0.88 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0.25 1 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5.2 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
GAD2-Sum, PCL2-Sum, Maslauch-9(Item4) predict Maslauch BURNOUT (For each participant across time)
mars = earth(PQL.Burnout ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=PQL.Burnout~Wellness.PHQ2.1+Wellness.PHQ2.1+Wellne…),
data=dw, pmethod=“backward”)
coefficients(Intercept) 29.00
h(1-Wellness.GAD2.2) -6.43
h(3-Wellness.Masl9.4) -2.17
h(Wellness.Masl9.9-4) -1.94
Selected 4 of 13 terms, and 3 of 17 predictors
Termination condition: GRSq -10 at 13 terms
Importance: Wellness.GAD2.2, Wellness.Masl9.4, Wellness.Masl9.9, …
Number of terms at each degree of interaction: 1 3 (additive model)
GCV 12.9 RSS 203 GRSq 0.567 RSq 0.738
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 0.88 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0.25 1 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5.2 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
SECTION III: CHOOSING THE BEST SCALES/ITEMS THAT PREDICT PERCEPTION (SELF, UNIT, & UVA) ACROSS TIME
PART 1: Choosing the Best Wellness Screen Items
Which wellness screen items predict perception of self, unit and UVA?
Maslauch9 (Item 9) predicts Current SELF Perception (For each participant across time)
mars = earth(Refl.Self ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=Refl.Self~Wellness.PHQ2.1+Wellness.PHQ2.1+Wellness…),
data=dw, pmethod=“backward”)
coefficients(Intercept) 1.723
h(Wellness.Masl9.9-4) -0.448
Selected 2 of 11 terms, and 1 of 17 predictors
Termination condition: GRSq -10 at 11 terms
Importance: Wellness.Masl9.9, Wellness.PHQ2.1-unused, …
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 0.781 RSS 17.4 GRSq 0.0261 RSq 0.165
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 0.88 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0.25 1 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5.2 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
Maslauch (Items 2 & 8) predict UNIT Perception (For each participant across time)
mars = earth(Refl.Unit ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=Refl.Unit~Wellness.PHQ2.1+Wellness.PHQ2.1+Wellness…),
data=dw, pmethod=“backward”)
coefficients(Intercept) 0.910
Wellness.Masl9.8 0.462
h(Wellness.Masl9.3-2) 0.249
Selected 3 of 12 terms, and 2 of 17 predictors
Termination condition: GRSq -10 at 12 terms
Importance: Wellness.Masl9.3, Wellness.Masl9.8, …
Number of terms at each degree of interaction: 1 2 (additive model)
GCV 0.343 RSS 6.49 GRSq 0.329 RSq 0.513
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 0.88 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0.25 1 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5.2 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
GAD2 (Item 1) predicts UVA Perception (For each participant across time)
mars = earth(Refl.UVA ~ Wellness.PHQ2.1 + Wellness.PHQ2.1 + Wellness.GAD2.1 +
Wellness.GAD2.2 + Wellness.PCL2.1 + Wellness.PCL2.2 + Wellness.AUDIT.1 +
Wellness.AUDIT.2 + Wellness.AUDIT.3 + Wellness.Masl9.1 + Wellness.Masl9.2 +
Wellness.Masl9.3 + Wellness.Masl9.4 + Wellness.Masl9.5 + Wellness.Masl9.6 +
Wellness.Masl9.7 + Wellness.Masl9.8 + Wellness.Masl9.9, data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=Refl.UVA~Wellness.PHQ2.1+Wellness.PHQ2.1+Wellness….),
data=dw, pmethod=“backward”)
coefficients(Intercept) 1.128
h(Wellness.GAD2.1-1) 0.544
Selected 2 of 13 terms, and 1 of 17 predictors
Termination condition: GRSq -Inf at 13 terms
Importance: Wellness.GAD2.1, Wellness.PHQ2.1-unused, …
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 0.419 RSS 9.35 GRSq 0.195 RSq 0.31
plotmo(mars)plotmo grid: Wellness.PHQ2.1 Wellness.GAD2.1 Wellness.GAD2.2
0 0.88 1
Wellness.PCL2.1 Wellness.PCL2.2 Wellness.AUDIT.1 Wellness.AUDIT.2
0.25 1 2 0
Wellness.AUDIT.3 Wellness.Masl9.1 Wellness.Masl9.2 Wellness.Masl9.3
0 5.2 0 2
Wellness.Masl9.4 Wellness.Masl9.5 Wellness.Masl9.6 Wellness.Masl9.7
2 0 5 1
Wellness.Masl9.8 Wellness.Masl9.9
0 5
PART 2: Choosing the Best PQL/PSS Composites
Which PQL/PSS scales best predict perception of Self, Unit and UVA?
No Full Scales Predict SELF Perception (For each participant across time)
mars = earth(Refl.Self ~ PQL.Compassion + PQL.Burnout + PQL.Stress + PSS.Stress,
data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=Refl.Self~PQL.Compassion+PQL.Burnout+PQL.Stress+PS…),
data=dw, pmethod=“backward”)
coefficients(Intercept) 1.37
Selected 1 of 11 terms, and 0 of 4 predictors
Termination condition: GRSq -10 at 11 terms
Importance: PQL.Compassion-unused, PQL.Burnout-unused, …
Number of terms at each degree of interaction: 1 (intercept only model)
GCV 0.802 RSS 20.9 GRSq 0 RSq 0
plotmo(mars)plotmo grid: PQL.Compassion PQL.Burnout PQL.Stress PSS.Stress
40 23 17 16
PQL-Burnout predicts UNIT Perception (For each participant across time)
mars = earth(Refl.Unit ~ PQL.Compassion + PQL.Burnout + PQL.Stress + PSS.Stress,
data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=Refl.Unit~PQL.Compassion+PQL.Burnout+PQL.Stress+PS…),
data=dw, pmethod=“backward”)
coefficients(Intercept) 1.61
h(24-PQL.Burnout) -0.11
Selected 2 of 11 terms, and 1 of 4 predictors
Termination condition: GRSq -10 at 11 terms
Importance: PQL.Burnout, PQL.Compassion-unused, PQL.Stress-unused, …
Number of terms at each degree of interaction: 1 1 (additive model)
GCV 0.429 RSS 9.58 GRSq 0.162 RSq 0.281
plotmo(mars)plotmo grid: PQL.Compassion PQL.Burnout PQL.Stress PSS.Stress
40 23 17 16
PQL-Burnout and PSS-Stress predict UVA Perception (For each participant across time)
mars = earth(Refl.UVA ~ PQL.Compassion + PQL.Burnout + PQL.Stress + PSS.Stress,
data = dw, pmethod = "backward")
summary(mars, digit = 3)Call:
earth(formula=Refl.UVA~PQL.Compassion+PQL.Burnout+PQL.Stress+PSS…),
data=dw, pmethod=“backward”)
coefficients(Intercept) 1.224
h(PQL.Burnout-24) -0.121
h(PSS.Stress-17) 0.138
Selected 3 of 10 terms, and 2 of 4 predictors
Termination condition: RSq changed by less than 0.001 at 10 terms
Importance: PSS.Stress, PQL.Burnout, PQL.Compassion-unused, …
Number of terms at each degree of interaction: 1 2 (additive model)
GCV 0.492 RSS 9.3 GRSq 0.0538 RSq 0.313
plotmo(mars)plotmo grid: PQL.Compassion PQL.Burnout PQL.Stress PSS.Stress
40 23 17 16
