In jonkeane/mocapGrip: Illusion-Grip Motion Capture Package

---

title: r params$title

knitr::opts_chunk$set(echo = FALSE)

Maximum grip aperture (on reach to grasp)

The maximum grip aperture is the maximum distance between the markers on the thumb and index finger during the period between when the hand left the table and when it touched the stick (this period is labeled as grip in our annotation system).

We fit a hierarchical (mixed effects) linear regression model.

The outcome (dependent) variable is:

• the maximum grip aperture (in mm)

The predictor (independent) variables are:

• the size of the stick in centimeters (centered at 8 cm, where 1 unit difference is 1 cm difference in stick size)
• the configuration of the fins (closed, none, open; where closed is the reference level)
• the interaction between the size of the stick and configuration of fins

There are varying intercepts by subject, and the effects of all the predictors (including interactions) are also allowed to vary by subject (these are the mixed effects). The full formula^[See the lme4 documentation (especially section 2) for explanation of the formula format.] for this model is:

maxGrip ~ stickcmCentered * fins + (1 + stickcmCentered * fins | obsisSubj)

Coefficient plot

These are the estimates of the effect sizes for the predictors (and interactions) in the model. The dot is the point estimate, the thick line is the 95% confidence interval, and the thin line the 99% confidence interval. One rule of thumb is: if the confidence intervals do not the effect is statistically significant.

mocapGrip:::CoefficientPlot(list(params$data[["action"]]$analyses[["maxGrip.stickAsContinuous"]]$bestModel[[1]]$modelObject))

Model predictions

These are predictions from the model for specific conditions. The dots are point estimates, and the lines are 95% confidence intervals.

ggplot(mocapGrip:::pred(params$data[["action"]]$analyses[["maxGrip.stickAsContinuous"]]$bestModel[[1]]$modelObject)) + aes(x=stickcmCentered+8, y=maxGrip, ymin=plo, ymax=phi, group=fins, color=fins) + geom_pointrange(position = position_dodge(width=0.5)) + labs(title = "Model predictions for grip", x="stick size (in cm)", y="the maximum grip aperture (in mm)")

Full model output

texreg::htmlreg(list(params$data[["action"]]$analyses[["maxGrip.stickAsContinuous"]]$bestModel[[1]]$modelObject),
        method = "boot", # only needed for the overriding of pvalues
        use.se=TRUE, # only needed for the overriding of pvalues
        #        override.pval = list(lengthLMscaleSum$coefficients[,"Pr(>|t|)"]),
        float.pos = "p!",
        single.row=TRUE,
        caption="Hierarchical linear regression coefficient estimates and standard errors.",
        use.packas=FALSE,
        custom.model.names=c("est. (s.e.)"),
        stars = c(0.001, 0.01, 0.05, 0.1),
        star.symbol = "\\*"
)

Intercept and slope adjustments

There are estimates of the intercept and slope adjustments by subject. and 95% confidence intervals around those estimates. This is also referred to as the mixed effects structure and is one way to see variability between subjects.

mocapGrip:::ggCaterpillar(lme4::ranef(params$data[["action"]]$analyses[["maxGrip.stickAsContinuous"]]$bestModel[[1]]$modelObject, condVar = TRUE))

Raw data

These are the raw data that was used in the model.

ggplot(params$data[["action"]]$data) + aes(x=stickcmCentered+8, y=maxGrip, group=fins, color=fins) + geom_point(position = position_dodge(width=0.5), alpha = 0.5) + labs(title = "Raw data for  grip", x="stick size (in cm)", y="the maximum grip aperture (in mm)")

Number of observations and occlusion

knitr::kable(
  params$data[["action"]]$data %>% dplyr::group_by(stick, fins) %>% dplyr::summarize(n=n(), subjs=length(unique(obsisSubj)), meanPerSubj=n/subjs)
)

if("action" %in% c("gestMove")){
  cat("#### Observations with gripType included *this is temporary*  \n The vast majority of the movements are open, with very few being open/closed. Currently, all of these (including the closed grips) are included in the analysis. We probably should exclude the closed, and possibly even the open.closed ones in the future. Thoughts?")

knitr::kable(
  params$data[["action"]]$data %>% dplyr::group_by(stick, fins, gripType) %>% dplyr::summarize(n=n(), subjs=length(unique(obsisSubj)), meanPerSubj=n/subjs)
)
}

There were r length(params$data[["action"]]$warnings) trials that were excluded because there was occlusion.

if(length(params$data[["action"]]$warnings)>0){
  # don't print anythign if there are no occlusions.
  cat(paste0("  \n* ", params$data[["action"]]$warnings, collapse = ""))
}

Size estimation

The mean grip aperture is the mean of the distance measurements between the markers on the thumb and index finger during the period when the subject said the word ready (this period is labeled as steady in our annotation system).

We fit a hierarchical (mixed effects) linear regression model.

The outcome (dependent) variable is:

• the mean grip aperture (in mm)

The predictor (independent) variables are:

• the size of the stick in centimeters (centered at 8 cm, where 1 unit difference is 1 cm difference in stick size)
• the configuration of the fins (closed, none, open; where closed is the reference level)
• the interaction between the size of the stick and configuration of fins

There are varying intercepts by subject, and the effects of all the predictors (including interactions) are also allowed to vary by subject (these are the mixed effects). The full formula^[See the lme4 documentation (especially section 2) for explanation of the formula format.] for this model is:

meanGrip ~ stickcmCentered * fins + (1 + stickcmCentered * fins | obsisSubj)

Coefficient plot

These are the estimates of the effect sizes for the predictors (and interactions) in the model. The dot is the point estimate, the thick line is the 95% confidence interval, and the thin line the 99% confidence interval. One rule of thumb is: if the confidence intervals do not the effect is statistically significant.

mocapGrip:::CoefficientPlot(list(params$data[["estimation"]]$analyses[["meanGrip.stickAsContinuous"]]$bestModel[[1]]$modelObject))

Model predictions

These are predictions from the model for specific conditions. The dots are point estimates, and the lines are 95% confidence intervals.

ggplot(mocapGrip:::pred(params$data[["estimation"]]$analyses[["meanGrip.stickAsContinuous"]]$bestModel[[1]]$modelObject)) + aes(x=stickcmCentered+8, y=meanGrip, ymin=plo, ymax=phi, group=fins, color=fins) + geom_pointrange(position = position_dodge(width=0.5)) + labs(title = "Model predictions for grip", x="stick size (in cm)", y="the mean grip aperture (in mm)")

Full model output

texreg::htmlreg(list(params$data[["estimation"]]$analyses[["meanGrip.stickAsContinuous"]]$bestModel[[1]]$modelObject),
        method = "boot", # only needed for the overriding of pvalues
        use.se=TRUE, # only needed for the overriding of pvalues
        #        override.pval = list(lengthLMscaleSum$coefficients[,"Pr(>|t|)"]),
        float.pos = "p!",
        single.row=TRUE,
        caption="Hierarchical linear regression coefficient estimates and standard errors.",
        use.packas=FALSE,
        custom.model.names=c("est. (s.e.)"),
        stars = c(0.001, 0.01, 0.05, 0.1),
        star.symbol = "\\*"
)

Intercept and slope adjustments

There are estimates of the intercept and slope adjustments by subject. and 95% confidence intervals around those estimates. This is also referred to as the mixed effects structure and is one way to see variability between subjects.

mocapGrip:::ggCaterpillar(lme4::ranef(params$data[["estimation"]]$analyses[["meanGrip.stickAsContinuous"]]$bestModel[[1]]$modelObject, condVar = TRUE))

Raw data

These are the raw data that was used in the model.

ggplot(params$data[["estimation"]]$data) + aes(x=stickcmCentered+8, y=meanGrip, group=fins, color=fins) + geom_point(position = position_dodge(width=0.5), alpha = 0.5) + labs(title = "Raw data for  grip", x="stick size (in cm)", y="the mean grip aperture (in mm)")

Number of observations and occlusion

knitr::kable(
  params$data[["estimation"]]$data %>% dplyr::group_by(stick, fins) %>% dplyr::summarize(n=n(), subjs=length(unique(obsisSubj)), meanPerSubj=n/subjs)
)

if("estimation" %in% c("gestMove")){
  cat("#### Observations with gripType included *this is temporary*  \n The vast majority of the movements are open, with very few being open/closed. Currently, all of these (including the closed grips) are included in the analysis. We probably should exclude the closed, and possibly even the open.closed ones in the future. Thoughts?")

knitr::kable(
  params$data[["estimation"]]$data %>% dplyr::group_by(stick, fins, gripType) %>% dplyr::summarize(n=n(), subjs=length(unique(obsisSubj)), meanPerSubj=n/subjs)
)
}

There were r length(params$data[["estimation"]]$warnings) trials that were excluded because there was occlusion.

if(length(params$data[["estimation"]]$warnings)>0){
  # don't print anythign if there are no occlusions.
  cat(paste0("  \n* ", params$data[["estimation"]]$warnings, collapse = ""))
}

jonkeane/mocapGrip documentation built on May 19, 2019, 7:30 p.m.