{\rtf1\ansi\ansicpg1252\cocoartf2580
\cocoatextscaling0\cocoaplatform0{\fonttbl\f0\fswiss\fcharset0 Helvetica;}
{\colortbl;\red255\green255\blue255;}
{\*\expandedcolortbl;;}
\margl1440\margr1440\vieww19040\viewh13280\viewkind0
\pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural\partightenfactor0
\f0\fs36 \cf0 \
AI_calibrations.RData contains two variables: AI_msd and Goals_Tasks_ID. AI_msd is a list of MSD calibrated item measures, rating category thresholds, and their standard errors, while Goals_Tasks_ID is a list of goal and task ID labels. Type any of the following commands into the command line to see what is stored in these variables:\
\
AI_msd$item_measures\
AI_msd$thresholds\
AI_msd$item_std_errors\
AI_msd$threshold_std_errors\
\
Goals_Tasks_ID$Goals\
Goals_Tasks_ID$Reading\
Goals_Tasks_ID$Mobility\
Goals_Tasks_ID$Vis_Info\
Goals_Tasks_ID$Vis_Motor\
Goals_Tasks_ID$All_Goals_Tasks\
\
The goal and task ID\'92s are explained in the two pdf files \'93Activity Inventory.pdf\'94 and Activity Inventory Instructions.pdf\'94\
\
The R package \'91msd\'92 must be installed to estimate person measures from the calibrated item measures and rating category thresholds. Documentation for \'91msd\'92 functions can be found here:\
https://www.rdocumentation.org/packages/msd\
\
We will use the pms function in the msd package to estimate person measures from calibrated item measures and rating category thresholds. \
\
For pms to work, the data type for the matrix with ordinal rating categories must be in the form of a \'93numeric matrix\'94 with 511 columns representing the 511 items in the AI, with the column order the same as in Goals_Tasks_ID$All_Goals_Tasks. The rows represent the persons. Each cell in the data matrix for the persons should be either an ordinal rating category or set to NA for missing data. It is important that there are no other types of entries such as a character (e.g., \'93m\'94 or \'93d\'94) nor should the cell be left blank, since R may record this as a different type of entry.\
\
Load your data file into the workspace using the following command (assume file name is \'93mydata.csv\'94):\
\
D = read.csv("mydata.csv", header = TRUE)\
\
The variable \'91D\'92 now contains the data. However, \'91D\'92 is a \'93data frame\'94 and not a numeric matrix as can be seen by typing\
\
class(D)\
\
Turn D into a numeric matrix using the following command:\
\
D2 = matrix(data = as.numeric(as.character(unlist(D))), nrow = dim(D)[1], ncol = dim(D)[2])\
\
Now the command class(D2) should say \'93matrix\'94 and D2 can be used in the pms function. Note that the above command to create D2 will work only if there are only ordinal rating categories and NA, and nothing else. Other characters may be represented with integers not in your data, and it is prudent to check if D2 looks the same as D. \
\
To estimate person measures with D2 and the calibrated item measures and thresholds, use the command:\
\
P = pms(data = D2, items = AI_msd$item_measures, thresholds = AI_msd$thresholds, misfit = TRUE, minRating = 0)\
\
The estimated parameters of P will be:\
\
P$person_measures\
P$person_std_errors\
P$infit_persons\
P$outfit_persons\
\
To use pms on a subset of D2, e.g., only the goals in the AI, the indexes of the columns for the subset of items must be known. For example, if only responses to the goals in the AI are to be considered, the indexes to the goals are:\
\
G_ind = match(Goals_Tasks_ID$Goals, Goals_Tasks_ID$All_Goals_Tasks)\
\
Now, pms can be used on just the goals as follows:\
\
P2 = pms(data = D2[,G_ind], items = AI_msd$item_measures[G_ind], thresholds = AI_msd$thresholds, misfit = TRUE)}
