1 Issues to work on
2 Dependent Variable: Functionality
- 2.1 Descriptive statistics
- 2.2 Regressions
  - 2.2.1 Logistic with selected variables (DV=functional)
  - 2.2.2 Logistic with all variables (DV=functional)
3 Dependent Variable: 3.3.a - length of last breakdown
- 3.1 Descriptive statistics
- 3.2 Regressions
4 Dependent Variable: 3.4.a - broken down in last year
5 Independent variables descriptive statistics

Link to this document: http://rpubs.com/anshulkumar/WHAgov1

1 Issues to work on

Many missing values are there in some of the regression variables, so very few of the observations are used. Maybe we can do a nested presentation.

2 Dependent Variable: Functionality

2.1 Descriptive statistics

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##   0.000   1.000   1.000   0.934   1.000   1.000      27

## db2$`_3_1_a_functional` 
##        n  missing distinct     Info      Sum     Mean      Gmd 
##      303       27        2    0.185      283    0.934   0.1237

Functionality
Functionality	Frequency		Total
Functionality	Count	NA	Total
0	20 6.1 %	0 0 %	20 6.1 %
1	283 85.8 %	0 0 %	283 85.8 %
NA	27 8.2 %	0 0 %	27 8.2 %
Total	330 100 %	0 100 %	330 100 %

2.1.1 Functionality and Water Source Type

sjt.xtab(db$Type,db$`_3_1_a_functional`,title = " ", show.row.prc = TRUE, show.na = TRUE, var.labels = c("Water Source Type", "Functionality"), show.summary = TRUE)


Water Source Type	Functionality			Total
Water Source Type	0	1	NA	Total
Borehole handpump	5 3.5 %	133 92.4 %	6 4.2 %	144 100 %
Borehole mech pump	0 0 %	22 100 %	0 0 %	22 100 %
Dug well open	3 6.5 %	43 93.5 %	0 0 %	46 100 %
Dug well handpump	13 18.3 %	56 78.9 %	2 2.8 %	71 100 %
Spring protected	3 7.5 %	35 87.5 %	2 5 %	40 100 %
Spring unprotected	0 0 %	10 100 %	0 0 %	10 100 %
Rain harvest tank	0 0 %	1 5.3 %	18 94.7 %	19 100 %
NA	0 0 %	0 0 %	2 100 %	2 100 %
Total	24 6.8 %	300 84.7 %	30 8.5 %	354 100 %
χ²=18.738 · df=6 · Cramer’s V=0.240 · Fisher’s p=0.009

2.1.2 Functionality and not-enough-funds

sjt.xtab(db$`_4_18_enoughfundsrepair`,db$`_3_1_a_functional`,title = " ", show.row.prc = TRUE, show.na = TRUE, var.labels = c("Enough Funds Repair", "Functionality"), show.summary = TRUE)


Enough Funds Repair	Functionality			Total
Enough Funds Repair	0	1	NA	Total
0	16 7.9 %	186 92.1 %	0 0 %	202 100 %
1	6 5.6 %	100 92.6 %	2 1.9 %	108 100 %
NA	2 4.5 %	14 31.8 %	28 63.6 %	44 100 %
Total	24 6.8 %	300 84.7 %	30 8.5 %	354 100 %
χ²=0.249 · df=1 · φ=0.042 · Fisher’s p=0.642

2.1.3 Functionality and fee collection system

sjt.xtab(db$`_4_4_feecollectionsystem`,db$`_3_1_a_functional`,title = " ", show.row.prc = TRUE, show.na = TRUE, var.labels = c("Fee collection system", "Functionality"), show.summary = TRUE)


Fee collection system	Functionality			Total
Fee collection system	0	1	NA	Total
0	15 8.8 %	153 90 %	2 1.2 %	170 100 %
1	7 4.7 %	143 95.3 %	0 0 %	150 100 %
NA	2 5.9 %	4 11.8 %	28 82.4 %	34 100 %
Total	24 6.8 %	300 84.7 %	30 8.5 %	354 100 %
χ²=1.622 · df=1 · φ=0.084 · p=0.203

2.1.4 Functionality and caretaker

sjt.xtab(db$`_2_2_a_Caretaker`,db$`_3_1_a_functional`,title = " ", show.row.prc = TRUE, show.na = TRUE, var.labels = c("Caretaker", "Functionality"), show.summary = TRUE)


Caretaker	Functionality			Total
Caretaker	0	1	NA	Total
0	17 9.3 %	164 89.6 %	2 1.1 %	183 100 %
1	7 5 %	133 95 %	0 0 %	140 100 %
NA	0 0 %	3 9.7 %	28 90.3 %	31 100 %
Total	24 6.8 %	300 84.7 %	30 8.5 %	354 100 %
χ²=1.612 · df=1 · φ=0.083 · p=0.204

2.1.5 Functionality and _4_17_planOandM

sjt.xtab(db$`_4_17_planOandM`,db$`_3_1_a_functional`,title = " ", show.row.prc = TRUE, show.na = TRUE, var.labels = c("Plan O and M", "Functionality"), show.summary = TRUE)


Plan O and M	Functionality			Total
Plan O and M	0	1	NA	Total
0	14 12.1 %	102 87.9 %	0 0 %	116 100 %
1	9 4.6 %	184 94.8 %	1 0.5 %	194 100 %
NA	1 2.3 %	14 31.8 %	29 65.9 %	44 100 %
Total	24 6.8 %	300 84.7 %	30 8.5 %	354 100 %
χ²=4.743 · df=1 · φ=0.137 · Fisher’s p=0.024

2.1.6 Functionality and Person Repair

sjt.xtab(db$`_4_19_personrepair`,db$`_3_1_a_functional`,title = " ", show.row.prc = TRUE, show.na = TRUE, var.labels = c("Person repair", "Functionality"), show.summary = TRUE)


Person repair	Functionality			Total
Person repair	0	1	NA	Total
0	18 8.6 %	190 90.9 %	1 0.5 %	209 100 %
1	5 4.7 %	101 94.4 %	1 0.9 %	107 100 %
NA	1 2.6 %	9 23.7 %	28 73.7 %	38 100 %
Total	24 6.8 %	300 84.7 %	30 8.5 %	354 100 %
χ²=1.076 · df=1 · φ=0.071 · Fisher’s p=0.256

2.2 Regressions

2.2.1 Logistic with selected variables (DV=functional)

# logistic _3_1_a_functional watersourcetype2 _2_2_a_Caretaker _2_3_inspectorvisits _3_1_d_primaryuse _4_4_feecollectionsystem _4_17_planOandM _4_18_enoughfundsrepair _4_19_personrepair _4_20_persontrained _5_0_latrineaccess _5_1_a_facilitator
# db$reportedfeesdividedbyallusers

tab_model(logit1 <- glm(`_3_1_a_functional` ~ Type+ `_2_2_a_Caretaker` +`_4_18_enoughfundsrepair` +`_4_19_personrepair`, data = db, family = "binomial"))

	3 1 a functional
Predictors	Odds Ratios	CI	p
(Intercept)	20.80	5.97 – 72.40	<0.001
Borehole mech pump	3568289.34	0.00 – Inf	0.991
Dug well open	0.63	0.12 – 3.35	0.587
Dug well handpump	0.14	0.04 – 0.48	0.002
Spring protected	0.42	0.08 – 2.16	0.301
Spring unprotected	4907073.21	0.00 – Inf	0.994
Rain harvest tank	4201877.71	0.00 – Inf	0.998
`_2_2_a_Caretaker`	1.71	0.62 – 4.70	0.297
`_4_18_enoughfundsrepair`	0.99	0.34 – 2.89	0.991
`_4_19_personrepair`	1.33	0.43 – 4.12	0.619
Observations	298
Cox & Snell’s R² / Nagelkerke’s R²	0.065 / 0.160

# did not work: 
# tab_model(logit1 <- glm(`_3_1_a_functional` ~ Type + `_2_2_a_Caretaker` + as.factor(`_3_1_d_primaryuse`) + as.factor(`_2_1_responsible`) +`_4_18_enoughfundsrepair`  +`_4_19_personrepair`, data = db, family = "binomial"))
# summary(logit1)

Variables _3_1_d_primaryuse and _2_1_responsible were omitted from this regression because they caused errors.

2.2.2 Logistic with all variables (DV=functional)

This model yields the error message we see below.

tab_model(glm(`_3_1_a_functional` ~ Type + `_2_2_a_Caretaker` +`_2_3_inspectorvisits` +as.factor(`_3_1_d_primaryuse`) +`_4_4_feecollectionsystem` +`_4_17_planOandM` +`_4_18_enoughfundsrepair` +`_4_19_personrepair` +`_4_20_persontrained` +`_5_0_latrineaccess` +`_5_1_a_facilitator`, data = db, family = "binomial"))

## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

	3 1 a functional
Predictors	Odds Ratios	CI	p
(Intercept)	70416446552692.53	0.00 – Inf	0.996
Borehole mech pump	217947894.36	0.00 – Inf	0.999
Dug well open	94912991.50	0.00 – Inf	0.999
Dug well handpump	0.06	0.00 – 4.26	0.199
Spring protected	1039086226.23	0.00 – Inf	0.999
Spring unprotected	108550026.68	0.00 – Inf	0.999
Rain harvest tank	3.16	0.00 – Inf	1.000
`_2_2_a_Caretaker`	0.07	0.00 – 9.74	0.287
`_2_3_inspectorvisits`	1.35	0.18 – 10.33	0.772
as factor(`3 1 d<br>primaryuse`)1	4.12	0.00 – Inf	1.000
as factor(`3 1 d<br>primaryuse`)2	0.03	0.00 – 3.45	0.143
as factor(`3 1 d<br>primaryuse`)3	96018860.80	0.00 – Inf	1.000
as factor(`3 1 d<br>primaryuse`)4	0.22	0.00 – 12.69	0.467
`_4_4_feecollectionsystem`	15.27	0.16 – 1429.15	0.239
`_4_17_planOandM`	0.00	0.00 – Inf	0.998
`_4_18_enoughfundsrepair`	0.50	0.01 – 18.10	0.705
`_4_19_personrepair`	3.78	0.12 – 119.10	0.450
`_4_20_persontrained`	75573443.09	0.00 – Inf	0.999
`_5_0_latrineaccess`	0.95	0.75 – 1.20	0.666
`_5_1_a_facilitator`	0.08	0.00 – 3.73	0.196
Observations	223
Cox & Snell’s R² / Nagelkerke’s R²	0.064 / 0.482

Cross-table of water source type and functionality:

with(db,table(Type,`_3_1_a_functional`, useNA = 'always'))

##                     _3_1_a_functional
## Type                   0   1 <NA>
##   Borehole handpump    5 133    6
##   Borehole mech pump   0  22    0
##   Dug well open        3  43    0
##   Dug well handpump   13  56    2
##   Spring protected     3  35    2
##   Spring unprotected   0  10    0
##   Rain harvest tank    0   1   18
##   <NA>                 0   0    2

Cross-table of fee collection systema and functionality:

with(db,table(`_4_4_feecollectionsystem`,`_3_1_a_functional`, useNA = 'always'))

##                         _3_1_a_functional
## _4_4_feecollectionsystem   0   1 <NA>
##                     0     15 153    2
##                     1      7 143    0
##                     <NA>   2   4   28

3 Dependent Variable: 3.3.a - length of last breakdown

3.1 Descriptive statistics

Length of breakdown:

Length of last breakdown
Length	Frequency		Total
Length	Count	NA	Total
0	1	0	1
0.5	6	0	6
0.75	3	0	3
1	37	0	37
2	29	0	29
3	3	0	3
3.5	1	0	1
4	11	0	11
5	2	0	2
7	15	0	15
10.5	1	0	1
11	1	0	1
12	1	0	1
14	9	0	9
15	2	0	2
16	2	0	2
20	2	0	2
21	2	0	2
28	1	0	1
30	10	0	10
31	1	0	1
38	1	0	1
45	2	0	2
60	6	0	6
61	1	0	1
65	1	0	1
84	1	0	1
90	3	0	3
100	1	0	1
120	6	0	6
150	2	0	2
180	1	0	1
360	3	0	3
365	2	0	2
420	1	0	1
1080	2	0	2
1440	1	0	1
1460	2	0	2
1800	1	0	1
3240	1	0	1
3600	2	0	2
6570	1	0	1
NA	173	0	173
Total	354	0	354

##   [1]   14.00    7.00   30.00    2.00      NA    1.00   30.00    1.00
##   [9]      NA   14.00      NA   28.00    1.00  100.00    1.00    2.00
##  [17]    1.00   30.00   84.00    2.00    0.50      NA      NA      NA
##  [25]   30.00   16.00   90.00    1.00    2.00      NA   11.00      NA
##  [33]    4.00   20.00   14.00    1.00    4.00    0.50    0.50    7.00
##  [41]      NA   14.00      NA      NA    4.00    2.00      NA      NA
##  [49]   30.00    1.00      NA   60.00  150.00    2.00 1080.00    1.00
##  [57]      NA    7.00    1.00   10.50    2.00  360.00   14.00    1.00
##  [65]    2.00      NA      NA   21.00    2.00    1.00   21.00   16.00
##  [73]    3.00    4.00   30.00  365.00    4.00    2.00      NA      NA
##  [81]      NA      NA      NA      NA      NA      NA    4.00      NA
##  [89]    2.00      NA      NA      NA      NA    4.00    4.00  150.00
##  [97]      NA      NA   60.00      NA      NA   15.00      NA    1.00
## [105]    2.00    1.00      NA      NA    7.00      NA      NA      NA
## [113]      NA      NA      NA      NA      NA      NA      NA    2.00
## [121]    7.00      NA      NA      NA      NA      NA    0.50      NA
## [129]    2.00    2.00      NA      NA   30.00      NA    5.00      NA
## [137]    0.00      NA    1.00      NA    2.00    3.00      NA    3.00
## [145]    1.00    1.00      NA      NA      NA    2.00   30.00    2.00
## [153]      NA      NA      NA    2.00      NA    3.50  365.00      NA
## [161]      NA    7.00   61.00    1.00      NA   14.00 1460.00    7.00
## [169]   38.00      NA      NA    4.00      NA      NA   30.00  120.00
## [177]      NA    4.00      NA 1800.00 3240.00      NA    1.00    2.00
## [185]      NA      NA      NA    2.00      NA      NA  120.00  420.00
## [193]      NA    1.00      NA      NA      NA      NA      NA    2.00
## [201]      NA      NA      NA      NA    2.00 1440.00      NA      NA
## [209]  180.00      NA      NA   60.00      NA      NA      NA      NA
## [217]      NA 1460.00      NA  360.00      NA      NA      NA      NA
## [225] 3600.00      NA      NA      NA   60.00 3600.00      NA      NA
## [233]      NA      NA      NA      NA      NA      NA   45.00      NA
## [241]   14.00      NA   12.00      NA      NA 1080.00      NA   90.00
## [249]  120.00      NA  120.00      NA      NA      NA      NA   30.00
## [257]    7.00      NA      NA   14.00      NA      NA      NA    2.00
## [265]      NA    4.00    1.00    1.00   90.00  120.00    0.75    1.00
## [273]    1.00  360.00      NA    7.00   45.00      NA   14.00    1.00
## [281]      NA    7.00    7.00    1.00    1.00   20.00      NA      NA
## [289]      NA    1.00    0.75   65.00   60.00 6570.00      NA    7.00
## [297]      NA      NA    1.00      NA   15.00    7.00    1.00    1.00
## [305]    2.00    2.00    1.00    1.00      NA    1.00    5.00      NA
## [313]  120.00    1.00   31.00    0.50    1.00    7.00    2.00    7.00
## [321]    0.75      NA      NA      NA      NA      NA      NA      NA
## [329]      NA    2.00      NA      NA      NA      NA      NA      NA
## [337]      NA      NA    1.00   60.00    0.50      NA      NA      NA
## [345]      NA      NA      NA      NA      NA      NA      NA      NA
## [353]      NA    2.00

Log transformed:

db$lengthbreakdown2 <- db$`_3_3_a_lengthbreakdown`
table(db$lengthbreakdown2, db$count, useNA = 'always')

##       
##        Count <NA>
##   0        1    0
##   0.5      6    0
##   0.75     3    0
##   1       37    0
##   2       29    0
##   3        3    0
##   3.5      1    0
##   4       11    0
##   5        2    0
##   7       15    0
##   10.5     1    0
##   11       1    0
##   12       1    0
##   14       9    0
##   15       2    0
##   16       2    0
##   20       2    0
##   21       2    0
##   28       1    0
##   30      10    0
##   31       1    0
##   38       1    0
##   45       2    0
##   60       6    0
##   61       1    0
##   65       1    0
##   84       1    0
##   90       3    0
##   100      1    0
##   120      6    0
##   150      2    0
##   180      1    0
##   360      3    0
##   365      2    0
##   420      1    0
##   1080     2    0
##   1440     1    0
##   1460     2    0
##   1800     1    0
##   3240     1    0
##   3600     2    0
##   6570     1    0
##   <NA>   173    0

db$lengthbreakdown3 <- recode(db$lengthbreakdown2,"0=NA")
table(db$lengthbreakdown3, db$count, useNA = 'always')

##       
##        Count <NA>
##   0.5      6    0
##   0.75     3    0
##   1       37    0
##   2       29    0
##   3        3    0
##   3.5      1    0
##   4       11    0
##   5        2    0
##   7       15    0
##   10.5     1    0
##   11       1    0
##   12       1    0
##   14       9    0
##   15       2    0
##   16       2    0
##   20       2    0
##   21       2    0
##   28       1    0
##   30      10    0
##   31       1    0
##   38       1    0
##   45       2    0
##   60       6    0
##   61       1    0
##   65       1    0
##   84       1    0
##   90       3    0
##   100      1    0
##   120      6    0
##   150      2    0
##   180      1    0
##   360      3    0
##   365      2    0
##   420      1    0
##   1080     2    0
##   1440     1    0
##   1460     2    0
##   1800     1    0
##   3240     1    0
##   3600     2    0
##   6570     1    0
##   <NA>   174    0

db$logLenBrk <- log(db$lengthbreakdown3)

sjp.frq(db$logLenBrk, type = "hist")

3.2 Regressions

3.2.1 OLS, without log transformation (DV = length of most recent breakdown)

# tab_model(ols0 <- lm(`_3_3_a_lengthbreakdown` ~ watersourcetype2 +`_2_2_a_Caretaker` +`_2_3_inspectorvisits` +`_3_1_d_primaryuse` +`_4_4_feecollectionsystem` +`_4_17_planOandM` +`_4_18_enoughfundsrepair` +`_4_19_personrepair` +`_4_20_persontrained` +`_5_0_latrineaccess` +`_5_1_a_facilitator` + `_4_9_actualmonthlyfees` + reportedfeesdividedbyallusers, data = db))

tab_model(ols0 <- lm(`_3_3_a_lengthbreakdown` ~ Type +`_2_2_a_Caretaker` + as.factor(`_3_1_d_primaryuse`) + as.factor(`_2_1_responsible`) +`_4_4_feecollectionsystem`+`_4_17_planOandM` +`_4_18_enoughfundsrepair` +`_4_19_personrepair`+`_4_20_persontrained`, data = db))

	3 3 a lengthbreakdown
Predictors	Estimates	CI	p
(Intercept)	175.74	-417.70 – 769.17	0.563
Borehole mech pump	-51.98	-530.26 – 426.30	0.832
Dug well open	-33.10	-753.66 – 687.46	0.928
Dug well handpump	-18.58	-280.39 – 243.24	0.890
Rain harvest tank	-267.91	-1528.12 – 992.29	0.678
`_2_2_a_Caretaker`	-133.83	-399.76 – 132.09	0.326
as factor(`3 1 d<br>primaryuse`)1	-181.09	-1456.87 – 1094.69	0.781
as factor(`3 1 d<br>primaryuse`)2	47.12	-255.68 – 349.93	0.761
as factor(`3 1 d<br>primaryuse`)4	69.13	-213.33 – 351.59	0.632
as factor(`2 1<br>responsible`)2	27.48	-511.75 – 566.72	0.921
as factor(`2 1<br>responsible`)4	-146.56	-1468.82 – 1175.69	0.828
as factor(`2 1<br>responsible`)5	-31.52	-1375.55 – 1312.51	0.963
as factor(`2 1<br>responsible`)6	97.12	-455.57 – 649.82	0.731
as factor(`2 1<br>responsible`)7	53.94	-1325.62 – 1433.49	0.939
as factor(`2 1<br>responsible`)8	-42.13	-757.38 – 673.12	0.908
as factor(`2 1<br>responsible`)9	66.91	-469.53 – 603.35	0.807
as factor(`2 1<br>responsible`)10	-49.69	-1062.06 – 962.68	0.924
`_4_4_feecollectionsystem`	-122.44	-393.44 – 148.56	0.378
`_4_17_planOandM`	139.32	-120.86 – 399.50	0.296
`_4_18_enoughfundsrepair`	30.33	-229.11 – 289.77	0.819
`_4_19_personrepair`	-194.83	-451.42 – 61.76	0.139
`_4_20_persontrained`	-16.69	-346.31 – 312.94	0.921
Observations	145
R² / adjusted R²	0.066 / -0.093

3.2.2 OLS, with log transformation (DV = log of length of most recent breakdown)

3.2.2.1 Model results

# tab_model(ols1 <- lm(logLenBrk ~ watersourcetype2 +`_2_2_a_Caretaker` +`_2_3_inspectorvisits` +`_3_1_d_primaryuse` +`_4_4_feecollectionsystem` +`_4_17_planOandM` +`_4_18_enoughfundsrepair` +`_4_19_personrepair` +`_4_20_persontrained` +`_5_0_latrineaccess` +`_5_1_a_facilitator` + `_4_9_actualmonthlyfees` + reportedfeesdividedbyallusers, data = db))

tab_model(ols1 <- lm(logLenBrk ~ Type +`_2_2_a_Caretaker` + as.factor(`_3_1_d_primaryuse`) + as.factor(`_2_1_responsible`) +`_4_4_feecollectionsystem`+`_4_17_planOandM` +`_4_18_enoughfundsrepair` +`_4_19_personrepair`+`_4_20_persontrained`, data = db))

	log Len Brk
Predictors	Estimates	CI	p
(Intercept)	2.31	0.56 – 4.05	0.011
Borehole mech pump	0.57	-0.85 – 1.98	0.435
Dug well open	1.52	-0.61 – 3.64	0.164
Dug well handpump	0.20	-0.58 – 0.99	0.611
Rain harvest tank	-1.02	-4.73 – 2.70	0.593
`_2_2_a_Caretaker`	-0.30	-1.09 – 0.48	0.448
as factor(`3 1 d<br>primaryuse`)1	-2.24	-6.03 – 1.54	0.248
as factor(`3 1 d<br>primaryuse`)2	-0.22	-1.11 – 0.68	0.637
as factor(`3 1 d<br>primaryuse`)4	0.69	-0.15 – 1.52	0.109
as factor(`2 1<br>responsible`)2	-0.16	-1.74 – 1.43	0.846
as factor(`2 1<br>responsible`)4	-0.62	-4.51 – 3.28	0.757
as factor(`2 1<br>responsible`)5	-0.68	-4.64 – 3.27	0.735
as factor(`2 1<br>responsible`)6	-0.04	-1.66 – 1.59	0.966
as factor(`2 1<br>responsible`)7	-1.75	-5.81 – 2.32	0.401
as factor(`2 1<br>responsible`)8	0.14	-1.97 – 2.24	0.899
as factor(`2 1<br>responsible`)9	-1.15	-2.73 – 0.42	0.155
as factor(`2 1<br>responsible`)10	-1.58	-4.56 – 1.40	0.301
`_4_4_feecollectionsystem`	-0.42	-1.23 – 0.38	0.305
`_4_17_planOandM`	0.87	0.10 – 1.64	0.029
`_4_18_enoughfundsrepair`	0.10	-0.66 – 0.87	0.790
`_4_19_personrepair`	-1.28	-2.05 – -0.50	0.002
`_4_20_persontrained`	0.18	-0.84 – 1.19	0.734
Observations	144
R² / adjusted R²	0.253 / 0.124

# fit is improved and observations are more when `_4_9_actualmonthlyfees` and reportedfeesdividedbyallusers are left out

3.2.2.2 Post-estimation tests

db$used <- TRUE
db$used[na.action(ols1)] <- FALSE

db$yols1 <- ifelse(db$used==TRUE, db$logLenBrk, NA)
describe(db$yols1)

## db$yols1 
##        n  missing distinct     Info     Mean      Gmd      .05      .10 
##      144      210       32    0.981    1.778     2.06  -0.2445   0.0000 
##      .25      .50      .75      .90      .95 
##   0.0000   1.3863   2.8284   4.3348   5.7548 
## 
## lowest : -0.6931472 -0.2876821  0.0000000  0.6931472  1.0986123
## highest:  5.8998974  6.0402547  7.2861917  7.4955419  8.7902691

yols1vec <- db$yols1
yols1a <- na.omit(yols1vec)
# describe(db$yols1)

# head(db[c("used","logLenBrk", "yols1")], n=50)
# with(db, table(used, useNA = 'always'))

ols1yhat <- fitted(ols1)
ols1rr <- resid(ols1, type = "response")
ols1rp <- resid(ols1, type = "pearson")

par(mfcol = c(2, 3))

# with(db, {
#   plot(ols1yhat, ols1rr, main = "Fitted vs Residuals")
#   qqnorm(ols1rr)
#   plot(ols1yhat, ols1rp, main = "Fitted vs Pearson Residuals")
#   qqnorm(ols1rp)
#   plot(TTC, ols1rp, main = "Actual vs Pearson Residuals")
#   plot(TTC, ols1yhat, main = "Actual vs Fitted")
# })

{
  plot(ols1yhat, ols1rr, main = "Fitted vs Residuals")
  qqnorm(ols1rr)
  plot(ols1yhat, ols1rp, main = "Fitted vs Pearson Residuals")
  qqnorm(ols1rp)
  plot(yols1a, ols1rp, main = "Actual vs Pearson Residuals")
  plot(yols1a, ols1yhat, main = "Actual vs Fitted")
}

hist(ols1rr, 
     main="Actual Residuals", 
     xlab="Residuals", 
     border="black", 
     col="skyblue",
     # xlim=c(0,125),
     las=1, 
     breaks=15)

# https://www.princeton.edu/~otorres/Regression101R.pdf # OLS diagnostics in r
# https://www.statmethods.net/stats/rdiagnostics.html # OLS diagnostics
# https://stats.stackexchange.com/questions/117052/replicating-statas-robust-option-in-r # robust standard errors
# https://data.princeton.edu/wws509/r/robust # robust 
residualPlots(ols1)

##                                Test stat Pr(>|Test stat|)  
## Type                                                       
## as.factor(`_3_1_d_primaryuse`)                             
## as.factor(`_2_1_responsible`)                              
## Tukey test                         1.885          0.05943 .
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

cor(yols1a,ols1yhat) # correlation of actual and fitted

## [1] 0.5029579

cor(ols1yhat, ols1rr) # correlation of predicted values and residuals

## [1] -6.592429e-17

3.2.3 OLS, with log transformation (DV = log of length of most recent breakdown), added variables

This model adds _4_9_actualmonthlyfees and reportedfeesdividedbyallusers, which reduces the sample size.

3.2.3.1 Model results - added variables

tab_model(ols1 <- lm(logLenBrk ~ Type +`_2_2_a_Caretaker` + as.factor(`_3_1_d_primaryuse`) + as.factor(`_2_1_responsible`) +`_4_4_feecollectionsystem`+`_4_17_planOandM` +`_4_18_enoughfundsrepair` +`_4_19_personrepair`+`_4_20_persontrained` + reportedfeesdividedbyallusers + `_4_9_actualmonthlyfees`, data = db))

	log Len Brk
Predictors	Estimates	CI	p
(Intercept)	0.94	-2.91 – 4.80	0.633
Borehole mech pump	0.03	-4.08 – 4.14	0.989
Dug well open	2.60	-0.81 – 6.02	0.141
Dug well handpump	-0.15	-1.13 – 0.82	0.759
`_2_2_a_Caretaker`	-0.38	-1.63 – 0.87	0.557
as factor(`3 1 d<br>primaryuse`)2	-0.61	-1.68 – 0.47	0.273
as factor(`3 1 d<br>primaryuse`)4	0.09	-0.83 – 1.01	0.849
as factor(`2 1<br>responsible`)2	-0.50	-2.19 – 1.19	0.563
as factor(`2 1<br>responsible`)6	-0.94	-2.86 – 0.98	0.343
as factor(`2 1<br>responsible`)8	-1.35	-4.41 – 1.70	0.389
as factor(`2 1<br>responsible`)9	-1.43	-3.23 – 0.38	0.126
as factor(`2 1<br>responsible`)10	-3.56	-7.26 – 0.14	0.064
`_4_4_feecollectionsystem`	1.63	-1.52 – 4.79	0.313
`_4_17_planOandM`	0.52	-0.40 – 1.44	0.271
`_4_18_enoughfundsrepair`	-0.22	-1.15 – 0.72	0.653
`_4_19_personrepair`	-0.47	-1.40 – 0.47	0.331
`_4_20_persontrained`	0.12	-1.34 – 1.59	0.872
reportedfeesdividedbyallusers	0.00	-0.00 – 0.00	0.442
`_4_9_actualmonthlyfees`	-0.00	-0.00 – 0.00	0.365
Observations	80
R² / adjusted R²	0.250 / 0.029

4 Dependent Variable: 3.4.a - broken down in last year

Coming soon

5 Independent variables descriptive statistics

# cor(db[, c('_2_2_a_Caretaker','_2_3_inspectorvisits','_3_1_d_primaryuse','_4_4_feecollectionsystem','_4_17_planOandM','_4_18_enoughfundsrepair','_4_19_personrepair','_4_20_persontrained','_5_0_latrineaccess','_5_1_a_facilitator')])

table(db$`_2_2_a_Caretaker`, db$count, useNA='always')

##       
##        Count <NA>
##   0      183    0
##   1      140    0
##   <NA>    31    0

table(db$`_2_3_inspectorvisits`, db$count, useNA='always')

##       
##        Count <NA>
##   0      222    0
##   0.5      4    0
##   1       48    0
##   2       13    0
##   3       10    0
##   4        5    0
##   6        1    0
##   9        2    0
##   12       1    0
##   120      1    0
##   <NA>    47    0

table(db$`_3_1_d_primaryuse`, db$count, useNA='always')

##       
##        Count <NA>
##   0      171    0
##   1        1    0
##   2       44    0
##   3        2    0
##   4       86    0
##   <NA>    50    0

table(db$`_4_4_feecollectionsystem`, db$count, useNA='always')

##       
##        Count <NA>
##   0      170    0
##   1      150    0
##   <NA>    34    0

table(db$`_4_17_planOandM`, db$count, useNA='always')

##       
##        Count <NA>
##   0      116    0
##   1      194    0
##   <NA>    44    0

table(db$`_4_18_enoughfundsrepair`, db$count, useNA='always')

##       
##        Count <NA>
##   0      202    0
##   1      108    0
##   <NA>    44    0

table(db$`_4_19_personrepair`, db$count, useNA='always')

##       
##        Count <NA>
##   0      209    0
##   1      107    0
##   <NA>    38    0

table(db$`_4_20_persontrained`, db$count, useNA='always')

##       
##        Count <NA>
##   0      279    0
##   1       41    0
##   <NA>    34    0

table(db$`_5_0_latrineaccess`, db$count, useNA='always')

##       
##        Count <NA>
##   2        1    0
##   5        1    0
##   30       1    0
##   50       3    0
##   55       1    0
##   70       8    0
##   75       7    0
##   77       1    0
##   77.5     1    0
##   80      22    0
##   82       1    0
##   85       8    0
##   87       1    0
##   90      26    0
##   92       2    0
##   93       3    0
##   95      20    0
##   96       1    0
##   97       4    0
##   97.5     2    0
##   98      19    0
##   99      16    0
##   99.5     2    0
##   100    166    0
##   <NA>    37    0

table(db$`_5_1_a_facilitator`, db$count, useNA='always')

##       
##        Count <NA>
##   0      181    0
##   1      107    0
##   <NA>    66    0

table(db$`_4_9_actualmonthlyfees`, db$count, useNA='always')

##         
##          Count <NA>
##   0          1    0
##   2700       1    0
##   5000       1    0
##   6000       1    0
##   10000      3    0
##   13000      1    0
##   15000      7    0
##   16000      2    0
##   18000      1    0
##   19000      1    0
##   20000      8    0
##   21000      1    0
##   22500      1    0
##   24000      2    0
##   25000      2    0
##   26000      1    0
##   27000      1    0
##   30000      6    0
##   35000      2    0
##   36000      1    0
##   40000      7    0
##   45000      7    0
##   46000      2    0
##   47500      1    0
##   48000      2    0
##   49000      1    0
##   50000      8    0
##   55000      2    0
##   60000      8    0
##   62500      1    0
##   65000      1    0
##   68000      1    0
##   70000      3    0
##   75000      3    0
##   80000      3    0
##   84000      1    0
##   85000      1    0
##   90000      2    0
##   96500      1    0
##   1e+05      5    0
##   106000     1    0
##   140000     1    0
##   141000     1    0
##   150000     2    0
##   160000     1    0
##   170000     1    0
##   192500     1    0
##   2e+05      2    0
##   220000     1    0
##   250000     1    0
##   280000     1    0
##   7e+05      1    0
##   9e+05      1    0
##   2e+06      1    0
##   <NA>     234    0

table(db$reportedfeesdividedbyallusers, db$count, useNA='always')

##                   
##                    Count <NA>
##   0                    1    0
##   30                   1    0
##   40                   1    0
##   51.2820510864258     1    0
##   60                   1    0
##   82.8729248046875     1    0
##   100                  2    0
##   112.5                1    0
##   125                  1    0
##   140                  1    0
##   153.846160888672     1    0
##   160                  1    0
##   171.428573608398     1    0
##   194.444442749023     1    0
##   200                  3    0
##   225                  3    0
##   230.769226074219     1    0
##   250                  3    0
##   300                  3    0
##   320                  1    0
##   327.868865966797     1    0
##   343.137268066406     1    0
##   350                  1    0
##   385.714294433594     1    0
##   400                  1    0
##   448.837219238281     1    0
##   450                  1    0
##   466.666656494141     1    0
##   500                  4    0
##   516.129028320312     1    0
##   571.428588867188     1    0
##   575                  1    0
##   588.235290527344     1    0
##   600                  1    0
##   617.9775390625       1    0
##   625                  2    0
##   642.857116699219     1    0
##   714.285705566406     1    0
##   722.222229003906     1    0
##   735.294128417969     1    0
##   800                  2    0
##   818.965515136719     1    0
##   833.333312988281     2    0
##   838.709655761719     1    0
##   900                  2    0
##   933.333312988281     1    0
##   941.176452636719     1    0
##   1000                 8    0
##   1007.14288330078     1    0
##   1080                 1    0
##   1088.88891601562     1    0
##   1111.11108398438     1    0
##   1145.83337402344     1    0
##   1200                 1    0
##   1203.125             1    0
##   1250                 4    0
##   1357.14282226562     1    0
##   1360                 1    0
##   1371.42858886719     1    0
##   1400                 1    0
##   1428.57141113281     1    0
##   1523.8095703125      1    0
##   1700                 1    0
##   1714.28576660156     1    0
##   1750                 1    0
##   1766.66662597656     1    0
##   1840                 1    0
##   1875                 1    0
##   2000                 2    0
##   2125                 1    0
##   2250                 1    0
##   2400                 1    0
##   2500                 2    0
##   2666.66674804688     2    0
##   3000                 1    0
##   3750                 1    0
##   4000                 1    0
##   4400                 1    0
##   7000                 1    0
##   10000                2    0
##   <NA>               244    0

table(db$`_2_1_responsible`, db$count, useNA='always')

##       
##        Count <NA>
##   1       12    0
##   2       70    0
##   3        8    0
##   4        1    0
##   5        5    0
##   6      119    0
##   7        2    0
##   8       63    0
##   9       70    0
##   10       4    0
##   <NA>     0    0

table(db$`__4_15_savingsstorage`, db$count, useNA='always')

##       
##        Count <NA>
##   0       42    0
##   1        4    0
##   3        2    0
##   4       37    0
##   5       56    0
##   6        9    0
##   <NA>   204    0

Regressions on governance variables

Prepared by Byron and Anshul

Version – 16 May 2019