]> code.communitydata.science - coldcallbot-discord.git/blob - assessment_and_tracking/compute_final_case_grades.R
rearrange the repository for publication
[coldcallbot-discord.git] / assessment_and_tracking / compute_final_case_grades.R
1 ## load in the data
2 #################################
3
4 case.sessions  <- 15
5 myuw <- read.csv("myuw-COM_482_A_autumn_2020_students.csv", stringsAsFactors=FALSE)
6
7 ## class-level variables
8 question.grades <- c("GOOD"=100, "FAIR"=100-(50/3.3), "BAD"=100-(50/(3.3)*2))
9 missed.question.penalty <- (50/3.3) * 0.2 ## 1/5 of a full point on the GPA scale
10
11 source("../assessment_and_tracking/track_participation.R")
12 setwd("case_grades")
13
14 rownames(d) <- d$discord.name
15
16 ## show the distribution of assessments
17 table(call.list.full$assessment)
18 prop.table(table(call.list.full$assessment))
19 table(call.list.full$answered)
20 prop.table(table(call.list.full$answered))
21
22 total.questions.asked <- nrow(call.list.full)
23
24 ## create new column with number of questions present
25 d$prop.asked <- d$num.calls / d$num.present
26
27 ## generate statistics using these new variables
28 prop.asks.quantiles <- quantile(d$prop.asked, probs=seq(0,1, 0.01))
29 prop.asks.quantiles <- prop.asks.quantiles[!duplicated(prop.asks.quantiles)]
30
31 ## this is generating broken stuff but it's not used for anything
32 d$prop.asked.quant <- cut(d$prop.asked, breaks=prop.asks.quantiles,
33     labels=names(prop.asks.quantiles)[1:(length(prop.asks.quantiles)-1)])
34
35 ## generate grades
36 ##########################################################
37
38 d$part.grade <- NA
39
40 ## print the median number of questions for (a) everybody and (b)
41 ## people that have been present 75% of the time
42 median(d$num.calls[d$days.absent < 0.25*case.sessions])
43 median(d$num.calls)
44
45 questions.cutoff <- median(d$num.calls)
46
47 ## helper function to generate average grade minus number of missing
48 gen.part.grade <- function (x.discord.name) {
49     q.scores <- question.grades[call.list$assessment[call.list$discord.name == x.discord.name]]
50     base.score <- mean(q.scores, na.rm=TRUE)
51
52     ## number of missing days
53     missing.days <- nrow(missing.in.class[missing.in.class$discord.name == x.discord.name,])
54
55     ## return the final score
56     data.frame(discord.name=x.discord.name,
57                part.grade=(base.score - missing.days * missed.question.penalty))
58 }
59
60 tmp <- do.call("rbind", lapply(d$discord.name[d$num.calls >= questions.cutoff], gen.part.grade))
61
62 d[as.character(tmp$discord.name), "part.grade"] <- tmp$part.grade
63
64 ## next handle the folks *under* the median
65
66 ## first we handle the zeros
67 ## step 1: first double check the people who have zeros and ensure that they didn't "just" get unlucky"
68 d[d$num.calls == 0,]
69
70 ## set those people to 0 :(
71 d$part.grade[d$num.calls == 0] <- 0
72
73 ## step 2 is to handle folks who got unlucky in the normal way
74 tmp <- do.call("rbind", lapply(d$discord.name[is.na(d$part.grade) & d$prop.asked <= median(d$prop.asked)], gen.part.grade))
75 d[as.character(tmp$discord.name), "part.grade"] <- tmp$part.grade
76
77 ## the people who are left are lucky and still undercounted so we'll penalize them
78 d[is.na(d$part.grade),]
79 penalized.discord.names <- d$discord.name[is.na(d$part.grade)]
80
81 ## generate the baseline participation grades as per the process above
82 tmp <- do.call("rbind", lapply(penalized.discord.names, gen.part.grade))
83 d[as.character(tmp$discord.name), "part.grade"] <- tmp$part.grade
84
85 ## now add "zeros" for every questions that is below the normal
86 d[as.character(penalized.discord.names),"part.grade"] <- ((
87     (questions.cutoff - d[as.character(penalized.discord.names),"num.calls"] * 0) +
88     (d[as.character(penalized.discord.names),"num.calls"] * d[as.character(penalized.discord.names),"part.grade"]) )
89     / questions.cutoff)
90
91 d[as.character(penalized.discord.names),]
92
93 ## map part grades back to 4.0 letter scale and points
94 d$part.4point <-round((d$part.grade / (50/3.3)) - 2.6, 2)
95
96 d[sort.list(d$prop.asked), c("discord.name", "num.calls", "num.present",
97                              "prop.asked", "prop.asked.quant", "part.grade", "part.4point",
98                              "days.absent")]
99
100 d[sort.list(d$part.4point), c("discord.name", "num.calls", "num.present",
101                              "prop.asked", "prop.asked.quant", "part.grade", "part.4point",
102                              "days.absent")]
103
104
105 ## writing out data
106 quantile(d$num.calls, probs=(0:100*0.01))
107 d.print <- merge(d, myuw[,c("StudentNo", "FirstName", "LastName", "UWNetID")],
108            by.x="student.num", by.y="StudentNo")
109 write.csv(d.print, file="final_participation_grades.csv")
110
111 library(rmarkdown)
112
113 for (x.discord.name in d$discord.name) {
114     render(input="../../assessment_and_tracking/student_report_template.Rmd",
115            output_format="html_document",
116            output_file=paste("../data/case_grades/student_reports/",
117                              d.print$UWNetID[d.print$discord.name == x.discord.name],
118                              sep=""))
119 }

Community Data Science Collective || Want to submit a patch?