Incivility Climate to Outcomes
Robert W. Walker
February 22, 2017
load("Ashley-Scaling-Complete.RData")A Summary of the Scaling
For each characteristic, we walked through this briefly but let me detail it here. For each scale created, I will show the basics of the fit. A single picture can describe it but the picture is not terrifically easy to read. The factor loadings can be represented by the slopes of the item characteristic curves. Let me first show civility climate. The direction of the latent variables and factors loading are sort of arbitrary, there is a postive and negative solution to every problem. So the way to read them is this.
The y axis gives the probability of a response of 3 or higher. Which actual number (1, 2, 3, … 6) chosen is arbitrary because the lines are parallel. There is information in them to compare (by where in the continuum they fall), but not that much. The x axis gives the latent factor as it increases from lowest to highest. The steeper the slope, the better the individual item does at differentiating those at higher than 2 [3–6] from those at 2 or below. Dashed-green [indicator 12] has the flattest slope; it does not respond to latent civility [or incivility as the case may be]. There is one case with two items that are opposite the other two. These are either reverse questions or something went wrong. This may need investigation.
The interpretation for each additional battery is the same except for the number of items. Sometimes more and sometimes less. Also note that I did not report the Injuries. They are binary batteries with 9 entities each.
Civility Climate
NB: 11 and 12 are dotted black and green respectively.
Burnout
Support: Organization
Something may be weird here.
Support: Managers
Support: Coworkers
Anxiety
Depression
Turnover Intentions
Continuance Commitment: Profession
Affective Commitment: Profession
Continuance Commitment
NB The 9.
Affective Commitment
Injuries are binaries. 9 in each case. I have not done anything with them yet.
Injuries: Month
Injuries: Week
Injuries
Perceived Power
Perceived Visibility
Irritation
Intensity
Intention
Workplace Aggression Frequency
Now I create the algorithm for estimating regressions on the factor score samples or estimating regressions with various factor score in generalized linear models with the individual forms of workplace aggression. A set of bivariate regressions or GLMs provides the first set of results. The interpretation, in each one, is z-scores. How often is a zero null hypothesis rejected? Or perhaps the p-values equivalently. They are calculate with either a GLM for the individual items, or a regression for the factor scores, over 10000 draws in each case where a factor is involved.
load("BivariateResults.RData")
summary(bivariate.res)## Estimate Std. Error t value Pr(>|t|)
## Min. :0.1432 Min. :0.04232 Min. :2.875 Min. :0.000e+00
## 1st Qu.:0.2450 1st Qu.:0.04965 1st Qu.:4.765 1st Qu.:3.800e-08
## Median :0.2674 Median :0.05154 Median :5.189 Median :3.570e-07
## Mean :0.2680 Mean :0.05162 Mean :5.197 Mean :1.121e-05
## 3rd Qu.:0.2905 3rd Qu.:0.05349 3rd Qu.:5.622 3rd Qu.:2.763e-06
## Max. :0.4114 Max. :0.06379 Max. :7.605 Max. :4.283e-03summary(VAgEx.res)## Estimate Std. Error z value Pr(>|z|)
## Min. :0.2756 Min. :0.08764 Min. :2.743 Min. :3.000e-09
## 1st Qu.:0.3909 1st Qu.:0.09781 1st Qu.:3.934 1st Qu.:6.754e-06
## Median :0.4238 Median :0.10038 Median :4.216 Median :2.491e-05
## Mean :0.4244 Mean :0.10052 Mean :4.218 Mean :9.699e-05
## 3rd Qu.:0.4566 3rd Qu.:0.10304 3rd Qu.:4.501 3rd Qu.:8.366e-05
## Max. :0.6284 Max. :0.11696 Max. :5.909 Max. :6.080e-03summary(IntimEx.res)## Estimate Std. Error z value Pr(>|z|)
## Min. :0.2682 Min. :0.09619 Min. :2.436 Min. :3.800e-08
## 1st Qu.:0.4011 1st Qu.:0.10762 1st Qu.:3.669 1st Qu.:2.718e-05
## Median :0.4343 Median :0.11049 Median :3.930 Median :8.501e-05
## Mean :0.4357 Mean :0.11066 Mean :3.934 Mean :2.502e-04
## 3rd Qu.:0.4691 3rd Qu.:0.11359 3rd Qu.:4.196 3rd Qu.:2.438e-04
## Max. :0.6583 Max. :0.12786 Max. :5.499 Max. :1.485e-02summary(ExclusEx.res)## Estimate Std. Error z value Pr(>|z|)
## Min. :0.2638 Min. :0.1088 Min. :2.205 Min. :7.400e-08
## 1st Qu.:0.4288 1st Qu.:0.1216 1st Qu.:3.476 1st Qu.:5.798e-05
## Median :0.4691 Median :0.1250 Median :3.752 Median :1.756e-04
## Mean :0.4699 Mean :0.1252 Mean :3.748 Mean :5.228e-04
## 3rd Qu.:0.5102 3rd Qu.:0.1286 3rd Qu.:4.021 3rd Qu.:5.081e-04
## Max. :0.7082 Max. :0.1467 Max. :5.382 Max. :2.745e-02summary(UnderEx.res)## Estimate Std. Error z value Pr(>|z|)
## Min. :0.2398 Min. :0.09956 Min. :2.290 Min. :5.300e-08
## 1st Qu.:0.4117 1st Qu.:0.11175 1st Qu.:3.627 1st Qu.:2.886e-05
## Median :0.4483 Median :0.11484 Median :3.908 Median :9.294e-05
## Mean :0.4496 Mean :0.11502 Mean :3.904 Mean :3.103e-04
## 3rd Qu.:0.4863 3rd Qu.:0.11808 3rd Qu.:4.182 3rd Qu.:2.870e-04
## Max. :0.6630 Max. :0.13429 Max. :5.441 Max. :2.202e-02summary(ICEx.res)## Estimate Std. Error z value Pr(>|z|)
## Min. :0.1803 Min. :0.1110 Min. :1.412 Min. :9.390e-06
## 1st Qu.:0.3452 1st Qu.:0.1253 1st Qu.:2.711 1st Qu.:1.095e-03
## Median :0.3844 Median :0.1288 Median :2.990 Median :2.787e-03
## Mean :0.3857 Mean :0.1290 Mean :2.985 Mean :5.691e-03
## 3rd Qu.:0.4253 3rd Qu.:0.1324 3rd Qu.:3.265 3rd Qu.:6.709e-03
## Max. :0.6146 Max. :0.1516 Max. :4.431 Max. :1.581e-01summary(PhAgEX.res)## Estimate Std. Error z value Pr(>|z|)
## Min. :-0.04844 Min. :0.1485 Min. :-0.2886 Min. :0.002073
## 1st Qu.: 0.23202 1st Qu.:0.1678 1st Qu.: 1.3588 1st Qu.:0.054213
## Median : 0.28452 Median :0.1725 Median : 1.6439 Median :0.100192
## Mean : 0.28458 Mean :0.1729 Mean : 1.6393 Mean :0.130354
## 3rd Qu.: 0.33654 3rd Qu.:0.1777 3rd Qu.: 1.9251 3rd Qu.:0.174203
## Max. : 0.55576 Max. :0.2044 Max. : 3.0796 Max. :0.850516And the algorithm.
vec10k <- seq(1,10000,by=1)
bivariate.res <- t(sapply(vec10k, function(x) { coefficients(summary(lm(WAF.Data.Scores[,x]~IC.Data.Scores[,x]*BO.Data.Scores[,x])))[c(2:4),]}))
VAgEx.res <- t(sapply(vec10k, function(x) {coefficients(summary(vglm(ordered(WAF.data.MCMC$VAgEx)~IC.Data.Scores[,x], family = cumulative(link = "logit", parallel = TRUE, reverse=TRUE), na.rm=TRUE)))[6,]}))
IntimEx.res <- t(sapply(vec10k, function(x) {coefficients(summary(vglm(ordered(WAF.data.MCMC$IntimEx)~IC.Data.Scores[,x], family = cumulative(link = "logit", parallel = TRUE, reverse=TRUE), na.rm=TRUE)))[6,]}))
ExclusEx.res <- t(sapply(vec10k, function(x) {coefficients(summary(vglm(ordered(WAF.data.MCMC$ExclusEx)~IC.Data.Scores[,x], family = cumulative(link = "logit", parallel = TRUE, reverse=TRUE), na.rm=TRUE)))[6,]}))
UnderEx.res <- t(sapply(vec10k, function(x) {coefficients(summary(vglm(ordered(WAF.data.MCMC$UnderEx)~IC.Data.Scores[,x], family = cumulative(link = "logit", parallel = TRUE, reverse=TRUE), na.rm=TRUE)))[6,]}))
ICEx.res <- t(sapply(vec10k, function(x) {coefficients(summary(vglm(ordered(WAF.data.MCMC$ICEx)~IC.Data.Scores[,x], family = cumulative(link = "logit", parallel = TRUE, reverse=TRUE), na.rm=TRUE)))[6,]}))
PhAgEX.res <- t(sapply(vec10k, function(x) {coefficients(summary(vglm(ordered(WAF.data.MCMC$PhAgEX)~IC.Data.Scores[,x], family = cumulative(link = "logit", parallel = TRUE, reverse=TRUE), na.rm=TRUE)))[5,]}))
Social Burden