Maggie Data

To summarize what we discussed, we are interested in knowing if the answers to questions 12 ("fellowship_yn"), 15 ("enfolded_postgrad_yn"), 16 ("private_academic"), 17 ("fellowship_years"), and 18 ("fellowship_field") differ based on year of training ("current_year"), debt ("debt"), gender ("gender"), race ("race"), age ("age"), marital status ("marital_status"), and family planning ("children").

Went from 265 to 257 rows because 8 cases from the original csv were removed due to having 50% or fewer of the questions answered.

read in and format data

df <- read.csv("maggie2.csv")
df <- df[!is.na(df$fellowship_yn),]


df2 <- df[c(2,7,20,22:23,31:33,35,37:39)]

df2$pgy_cat <- ifelse(df2$current_year==1|df2$current_year==2, 'junior', ifelse(df2$current_year==7|df2$current_year==6, "senior", "midlevel"))

df2$fellow_cat <- ifelse(df2$fellowship_yn==1|df2$fellowship_yn==2, 'Probably or Definitely Yes', ifelse(df2$fellowship_yn==4|df2$fellowship_yn==5, "Probably or Definitely No", "Undecided"))

df2$age_cat <- ifelse(df2$age==2, '22-25 y/o', ifelse(df2$age==3, "26-30 y/o", ifelse(df2$age==4, "31-35 y/o", "36-45 y/o")))

df2$race_cat <- ifelse(df2$race==1, 'Native American or Alaska Native', ifelse(df2$race==2, "Asian", ifelse(df2$race==3, "Black or African American", ifelse(df2$race==4, "Native Hawaiian or other Pacific Islander", ifelse(df2$race==5, "White", "Other/No Response")))))

df2$gender_cat <- ifelse(df2$gender==1, 'Cis Woman', ifelse(df2$gender==2, "Cis Man", ifelse(df2$gender==3, "NB", ifelse(df2$gender==4, "Trans Woman", ifelse(df2$gender==7, "No Response", "Other")))))

RACE x fellowship

1. native american/alaskan
2. asian
3. black/AA
4. native hawaiin or pacific islander
5. white

df_race <- df2[!df2$race==7,]
df_race <- df_race[!df_race$race==6,]
df_race <- df_race[!is.na(df_race$fellowship_yn),]

race <- table(df_race$fellow_cat,df_race$race_cat)
race2 <- as.data.frame(table(df_race$fellow_cat,df_race$race_cat))
colnames(race2) <- c("fellow_cat","race","count")

race_pct <- setDT(race2)[, list(Sum_Count = sum(count)), keyby = list(race, fellow_cat)][, 
            Count_Pct := round(Sum_Count/sum(Sum_Count), 2)*100, by = race][]
race_pct2 <- setDT(race2)[, list(sum_count = sum(count)), keyby = list(fellow_cat,race)][, 
            pct := round(sum_count/sum(sum_count), 2)*100, by = fellow_cat][]


p_all<-ggplot(data=race_pct, aes(x=race, y=Count_Pct, fill=fellow_cat))  + geom_bar(stat="identity", color="black", position=position_dodge())+
  theme_minimal()+ scale_fill_brewer(palette="Blues")+gghisto+ggtitle("fellowship committments for all races") + #scale_x_discrete(labels=c("NA or Alaskan","Asian","Black or AA", "Hawaiian or PI","White")) + 
  ylim(0,100)+
     geom_text(aes(label=Count_Pct), position=position_dodge(width=0.9), vjust=-0.25)+ theme(axis.text.x = element_text(size=6, angle=15))
p_all

fishers tests between groups race and fellow cat

fisher.test(df_race$race_cat,df_race$fellow_cat, simulate.p.value = TRUE, B=2000)

    Fisher's Exact Test for Count Data with simulated p-value (based on 2000 replicates)

data:  df_race$race_cat and df_race$fellow_cat
p-value = 0.0004998
alternative hypothesis: two.sided

fisher.test(table(df_race$fellow_cat,df_race$race_cat), simulate.p.value = TRUE, B=2000)

    Fisher's Exact Test for Count Data with simulated p-value (based on 2000 replicates)

data:  table(df_race$fellow_cat, df_race$race_cat)
p-value = 0.0004998
alternative hypothesis: two.sided

Conclusion from fisher’s test: there are significant differences between race and fellowship plans.

Below: 2-group fishers tests

Native American excluded for low sample size (n=4)

white.asian <- table(df_race[df_race$race_cat=="White" | df_race$race_cat=="Asian",14],df_race[df_race$race_cat=="White" | df_race$race_cat=="Asian",16])

white.black <- table(df_race[df_race$race_cat=="White" | df_race$race_cat=="Black or African American",14],df_race[df_race$race_cat=="White" | df_race$race_cat=="Black or African American",16])

white.NHPA <- table(df_race[df_race$race_cat=="White" | df_race$race_cat=="Native Hawaiian or other Pacific Islander",14],df_race[df_race$race_cat=="White" | df_race$race_cat=="Native Hawaiian or other Pacific Islander",16])

asian.black <- table(df_race[df_race$race_cat=="Asian" | df_race$race_cat=="Black or African American",14],df_race[df_race$race_cat=="Asian" | df_race$race_cat=="Black or African American",16])

asian.NHPA <- table(df_race[df_race$race_cat=="Asian" | df_race$race_cat=="Native Hawaiian or other Pacific Islander",14],df_race[df_race$race_cat=="Asian" | df_race$race_cat=="Native Hawaiian or other Pacific Islander",16])

black.NHPA <- table(df_race[df_race$race_cat=="Black or African American" | df_race$race_cat=="Native Hawaiian or other Pacific Islander",14],df_race[df_race$race_cat=="Black or African American" | df_race$race_cat=="Native Hawaiian or other Pacific Islander",16])

#tests
fisher.test(white.asian, simulate.p.value = TRUE, B=2000) #S

    Fisher's Exact Test for Count Data with simulated p-value (based on 2000 replicates)

data:  white.asian
p-value = 0.0004998
alternative hypothesis: two.sided

fisher.test(white.black, simulate.p.value = TRUE, B=2000) #NS

    Fisher's Exact Test for Count Data with simulated p-value (based on 2000 replicates)

data:  white.black
p-value = 0.07746
alternative hypothesis: two.sided

fisher.test(white.NHPA, simulate.p.value = TRUE, B=2000) #S

    Fisher's Exact Test for Count Data with simulated p-value (based on 2000 replicates)

data:  white.NHPA
p-value = 0.0009995
alternative hypothesis: two.sided

fisher.test(asian.black, simulate.p.value = TRUE, B=2000) #S -> NS after corrections. 

    Fisher's Exact Test for Count Data with simulated p-value (based on 2000 replicates)

data:  asian.black
p-value = 0.01549
alternative hypothesis: two.sided

fisher.test(asian.NHPA, simulate.p.value = TRUE, B=2000) #NS

    Fisher's Exact Test for Count Data with simulated p-value (based on 2000 replicates)

data:  asian.NHPA
p-value = 1
alternative hypothesis: two.sided

fisher.test(black.NHPA, simulate.p.value = TRUE, B=2000) #NS

    Fisher's Exact Test for Count Data

data:  black.NHPA
p-value = 0.09753
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
  0.6423599 35.5109712
sample estimates:
odds ratio 
  4.298978 

df_race[,13:16] %>%
  tbl_summary(by=race_cat) %>%
  add_p()

There was an error in 'add_p()/add_difference()' for variable 'fellow_cat', p-value omitted:
Error in stats::fisher.test(c("Probably or Definitely No", "Probably or Definitely Yes", : FEXACT error 7(location). LDSTP=18630 is too small for this problem,
  (pastp=19.1276, ipn_0:=ipoin[itp=423]=279, stp[ipn_0]=16.7221).
Increase workspace or consider using 'simulate.p.value=TRUE'
There was an error in 'add_p()/add_difference()' for variable 'age_cat', p-value omitted:
Error in stats::fisher.test(c("36-45 y/o", "26-30 y/o", "36-45 y/o", "31-35 y/o", : FEXACT error 6.  LDKEY=621 is too small for this problem,
  (ii := key2[itp=295] = 3790869, ldstp=18630)
Try increasing the size of the workspace and possibly 'mult'

Characteristic	Asian, N = 321	Black or African American, N = 221	Native American or Alaska Native, N = 41	Native Hawaiian or other Pacific Islander, N = 121	White, N = 1541	p-value2
pgy_cat						0.053
junior	5 (16%)	7 (33%)	0 (0%)	7 (58%)	34 (22%)
midlevel	18 (56%)	12 (57%)	4 (100%)	3 (25%)	96 (62%)
senior	9 (28%)	2 (9.5%)	0 (0%)	2 (17%)	24 (16%)
Unknown	0	1	0	0	0
fellow_cat
Probably or Definitely No	1 (3.1%)	0 (0%)	0 (0%)	0 (0%)	8 (5.2%)
Probably or Definitely Yes	17 (53%)	19 (86%)	1 (25%)	7 (58%)	141 (92%)
Undecided	14 (44%)	3 (14%)	3 (75%)	5 (42%)	5 (3.2%)
age_cat
22-25 y/o	4 (12%)	2 (9.1%)	0 (0%)	3 (25%)	18 (12%)
26-30 y/o	8 (25%)	8 (36%)	1 (25%)	5 (42%)	76 (49%)
31-35 y/o	16 (50%)	11 (50%)	0 (0%)	4 (33%)	50 (32%)
36-45 y/o	4 (12%)	1 (4.5%)	3 (75%)	0 (0%)	10 (6.5%)
1 n (%)
2 Fisher's exact test

Native American or Alaskan: Prob/Def No 0% Prob/Def Yes 25% Unsure 75%

Asian Prob/Def No 3% Prob/Def Yes 53% Unsure 44%

Black or African American Prob/Def No 0% Prob/Def Yes 86% Unsure 14%

Native Hawaiian or Pacific Islander Prob/Def No 0% Prob/Def Yes 58% Unsure 42%

White Prob/Def No 5% Prob/Def Yes 92% Unsure 3%

GENDER x fellowship

1. cis woman
2. cis man
3. NB
4. T woman
5. T man
6. Not Listed
7. No response

df_gen <- df2[!df2$gender==5,]
df_gen <- df_gen[!df_gen$gender==6,]
df_gen <- df_gen[!is.na(df_gen$fellowship_yn),]

#df_gen2 is only cis men, cis women, and NR
df_gen2 <- df_gen[!df_gen$gender==3,]
df_gen2 <- df_gen2[!df_gen2$gender==4,]

gender <- table(df_gen$fellow_cat,df_gen$gender_cat)
gender2 <- as.data.frame(table(df_gen$fellow_cat,df_gen$gender_cat))
colnames(gender2) <- c("fellow_cat","gender","count")

gen_pct <- setDT(gender2)[, list(sum_count = sum(count)), keyby = list(fellow_cat,gender)][, 
            pct := round(sum_count/sum(sum_count), 2)*100, by = fellow_cat][]
gen_pct2 <- setDT(gender2)[, list(sum_count = sum(count)), keyby = list(fellow_cat,gender)][, 
            pct := round(sum_count/sum(sum_count), 2)*100, by = gender][]


p_all<-ggplot(data=gen_pct2, aes(x=gender, y=pct, fill=fellow_cat))  + geom_bar(stat="identity", color="black", position=position_dodge())+
  theme_minimal()+ scale_fill_brewer(palette="Blues")+gghisto+ggtitle("fellowship committments for all genders") + #scale_x_discrete(labels=c("Cis Woman","Cis Man","NB", "T Woman","No Response")) + 
  ylim(0,100)+
     geom_text(aes(label=pct), position=position_dodge(width=0.9), vjust=-0.25)+ theme(axis.text.x = element_text(size=12, angle=15))
p_all

The categories for TW and NB are not normally distributed compared to the other 3 categories. “In 2018, 0.7% of matriculating medical students identified as TGNB” https://www.aamc.org/data-reports/students-residents/report/matriculating-student-questionnaire-msq

df_gen[,13:17] %>%
  tbl_summary(by=gender_cat)%>%
  add_p()

There was an error in 'add_p()/add_difference()' for variable 'age_cat', p-value omitted:
Error in stats::fisher.test(c("36-45 y/o", "26-30 y/o", "26-30 y/o", "36-45 y/o", : FEXACT error 6.  LDKEY=621 is too small for this problem,
  (ii := key2[itp=562] = 5767166, ldstp=18630)
Try increasing the size of the workspace and possibly 'mult'
There was an error in 'add_p()/add_difference()' for variable 'race_cat', p-value omitted:
Error in stats::fisher.test(c("White", "Other/No Response", "White", "White", : FEXACT[f3xact()] error: hash key 7e+09 > INT_MAX, kyy=153, it[i (= nco = 6)]= -13532483.
Rather set 'simulate.p.value=TRUE'

Characteristic	Cis Man, N = 1511	Cis Woman, N = 591	NB, N = 81	No Response, N = 111	Trans Woman, N = 151	p-value2
pgy_cat						0.048
junior	26 (17%)	20 (35%)	4 (50%)	1 (20%)	5 (33%)
midlevel	95 (63%)	28 (49%)	2 (25%)	2 (40%)	8 (53%)
senior	30 (20%)	9 (16%)	2 (25%)	2 (40%)	2 (13%)
Unknown	0	2	0	6	0
fellow_cat						<0.001
Probably or Definitely No	5 (3.3%)	4 (6.8%)	0 (0%)	1 (9.1%)	0 (0%)
Probably or Definitely Yes	133 (88%)	51 (86%)	5 (62%)	9 (82%)	6 (40%)
Undecided	13 (8.6%)	4 (6.8%)	3 (38%)	1 (9.1%)	9 (60%)
age_cat
22-25 y/o	18 (12%)	4 (7.0%)	0 (0%)	0 (0%)	1 (6.7%)
26-30 y/o	61 (40%)	32 (56%)	2 (25%)	3 (27%)	5 (33%)
31-35 y/o	57 (38%)	16 (28%)	5 (62%)	5 (45%)	7 (47%)
36-45 y/o	15 (9.9%)	5 (8.8%)	1 (12%)	3 (27%)	2 (13%)
Unknown	0	2	0	0	0
race_cat
Asian	15 (9.9%)	6 (10%)	0 (0%)	0 (0%)	9 (60%)
Black or African American	15 (9.9%)	1 (1.7%)	1 (12%)	1 (9.1%)	2 (13%)
Native American or Alaska Native	4 (2.6%)	0 (0%)	0 (0%)	0 (0%)	0 (0%)
Native Hawaiian or other Pacific Islander	3 (2.0%)	0 (0%)	6 (75%)	0 (0%)	3 (20%)
Other/No Response	11 (7.3%)	4 (6.9%)	0 (0%)	10 (91%)	0 (0%)
White	103 (68%)	47 (81%)	1 (12%)	0 (0%)	1 (6.7%)
Unknown	0	1	0	0	0
1 n (%)
2 Fisher's exact test

Redoing the analysis and excluding NB and NR gender categories

gender2 <- as.data.frame(table(df_gen2$fellow_cat,df_gen2$gender_cat))
colnames(gender2) <- c("fellow_cat","gender","count")

gen_pct <- setDT(gender2)[, list(sum_count = sum(count)), keyby = list(fellow_cat,gender)][, 
            pct := round(sum_count/sum(sum_count), 2)*100, by = fellow_cat][]
gen_pct2 <- setDT(gender2)[, list(sum_count = sum(count)), keyby = list(fellow_cat,gender)][, 
            pct := round(sum_count/sum(sum_count), 2)*100, by = gender][]


p_all<-ggplot(data=gen_pct2, aes(x=gender, y=pct, fill=fellow_cat))  + geom_bar(stat="identity", color="black", position=position_dodge())+
  theme_minimal()+ scale_fill_brewer(palette="Blues")+gghisto+ggtitle("fellowship committments for all genders") + ylim(0,100)+
     geom_text(aes(label=pct), position=position_dodge(width=0.9), vjust=-0.25)
p_all

fishers tests between groups race and fellow cat

fisher.test(df_gen2$gender_cat,df_gen2$fellow_cat, simulate.p.value = TRUE, B=2000)

    Fisher's Exact Test for Count Data with simulated p-value (based on 2000 replicates)

data:  df_gen2$gender_cat and df_gen2$fellow_cat
p-value = 0.4973
alternative hypothesis: two.sided

fisher.test(table(df_gen2$fellow_cat,df_gen2$gender_cat), simulate.p.value = TRUE, B=2000)

    Fisher's Exact Test for Count Data with simulated p-value (based on 2000 replicates)

data:  table(df_gen2$fellow_cat, df_gen2$gender_cat)
p-value = 0.5012
alternative hypothesis: two.sided

df_gen2[,13:17] %>%
  tbl_summary(by=gender_cat)%>%
  add_p()

There was an error in 'add_p()/add_difference()' for variable 'race_cat', p-value omitted:
Error in stats::fisher.test(c("White", "Other/No Response", "White", "White", : FEXACT error 7(location). LDSTP=18630 is too small for this problem,
  (pastp=26.1932, ipn_0:=ipoin[itp=38]=1202, stp[ipn_0]=28.4756).
Increase workspace or consider using 'simulate.p.value=TRUE'

Characteristic	Cis Man, N = 1511	Cis Woman, N = 591	No Response, N = 111	p-value2
pgy_cat				0.043
junior	26 (17%)	20 (35%)	1 (20%)
midlevel	95 (63%)	28 (49%)	2 (40%)
senior	30 (20%)	9 (16%)	2 (40%)
Unknown	0	2	6
fellow_cat				0.5
Probably or Definitely No	5 (3.3%)	4 (6.8%)	1 (9.1%)
Probably or Definitely Yes	133 (88%)	51 (86%)	9 (82%)
Undecided	13 (8.6%)	4 (6.8%)	1 (9.1%)
age_cat				0.2
22-25 y/o	18 (12%)	4 (7.0%)	0 (0%)
26-30 y/o	61 (40%)	32 (56%)	3 (27%)
31-35 y/o	57 (38%)	16 (28%)	5 (45%)
36-45 y/o	15 (9.9%)	5 (8.8%)	3 (27%)
Unknown	0	2	0
race_cat
Asian	15 (9.9%)	6 (10%)	0 (0%)
Black or African American	15 (9.9%)	1 (1.7%)	1 (9.1%)
Native American or Alaska Native	4 (2.6%)	0 (0%)	0 (0%)
Native Hawaiian or other Pacific Islander	3 (2.0%)	0 (0%)	0 (0%)
Other/No Response	11 (7.3%)	4 (6.9%)	10 (91%)
White	103 (68%)	47 (81%)	0 (0%)
Unknown	0	1	0
1 n (%)
2 Fisher's exact test