Flagging is a way of identifying records for which there is a strong likelihood that values of anthropometric measurements or the age of the child are incorrect. Records can then be checked and corrected, or censored (i.e. excluded) from subsequent analyses.
Flagging is a process of checking whether values of anthropometric indices are outside a given range and recording the result in one or more new variables. The result may be a set of logical (i.e. 1/0 or true/false) flag variables (i.e. one flag variable per anthropometric index) or a single variable holding a code number that classifies the nature of the detected problem(s).
Flagging is usually applied to height-for-age z-scores (HAZ), weight-for-age z-scores (WAZ), weight-for-height z-scores (WHZ), and BMI-for-age z-scores (BAZ) calculated from data collected during nutritional anthropometry surveys. The flagging process can be easily applied to other variables.
Two flagging criteria for anthropometric indices are in common use internationally. These are the WHO flagging criteria and the SMART flagging criteria. Both methods flag records in which one or more anthropometric indices are more than a certain distance either side of a reference value. The two methods are summarised in table below.
| Lower limit | Upper limit | Lower limit | Upper limit | Lower limit | Upper limit | Lower limit | Upper limit | Reference value | |
|---|---|---|---|---|---|---|---|---|---|
| WHO | -6 | +6 | -6 | +5 | -5 | +5 | -5 | +5 | Zero |
| SMART | -3 | +3 | -3 | +3 | -3 | +3 | NA\(^ 2\) | NA\(^ 2\) | Survey sample |
| 1 Indices are height-for-age z-score (HAZ), weight-for-age z-score (WAZ), weight-for-height z-score (WHZ), and BMI-for-age z-score (BAZ). | |||||||||
| 2 NA = Not available. BAZ is not used in SMART surveys. SMART flagging criteria for BAZ are undefined. |
Applying flagging criteria is a matter of checking that individual values of these indices are within the lower and upper limits shown in the table above. Values that are outside of these limits are flagged in a new variable.
The WHO criteria are simple biologically plausible ranges. If, for example, a value for WHZ is below -5 or above +5 then the record is flagged to indicate a likely problem with WHZ. This will usually be caused by an erroneous weight or height value being recorded.
Note that values outside of these flagging limits may be observed in children admitted into (e.g.) therapeutic feeding programs.
SMART criteria are more complicated. They require the mean value of the index to be calculated from the survey data. This is then used as the reference value. For example, if a value for WHZ is below:
\[ \text{mean WHZ} ~ - ~ 3 \]
or above:
\[ \text{mean WHZ} ~ + ~ 3 \]
then the record is flagged to indicate a likely problem with WHZ.
A mean WHZ of -1.15 gives lower and upper SMART flagging limits of:
\[ -1.15 ~ - ~ 3 ~ = ~ -4.15 \]
and:
\[ -1.15 ~ + ~ 3 ~ = ~ +1.85 \]
respectively. These limits may incorrectly flag biologically plausible values. See figure below.
Example of WHO and SMART flagging criteria for weight-for-height z-scores (WHZ)
The WHO and SMART flagging criteria will flag different but overlapping sets of measurements. This means that survey results can be affected by the flagging criteria used. This is because the prevalence of an indicator describes the proportion of values in one of the “tails” of the distribution of an index (see figure below).
The SMART flagging criteria will usually flag more records than the WHO flagging criteria. This will act to reduce estimated prevalence (see figure below). This will be a particular problem when the prevalence of severe forms of undernutrition is high.
There are some problems with using the SMART flagging criteria:
Flagging is about detecting outlier values. The SMART flagging criteria use distance from the sample mean, but the value of the mean can be strongly influenced by the presence of outliers. This could be overcome by, for example, using the median or a trimmed mean as the reference mean. If you do this you will not be using the SMART flagging criteria.
SMART flagging criteria are supposed to define outliers using statistically plausible limits. The underlying principle is that, for a normally distributed variable, we expect 99.87% of all values to lie within three sample standard deviations of the sample mean. If we exclude records with values more than three standard deviations from the mean then we would incorrectly flag very few records (i.e. 0.13% of the total) as problematic. The SMART method assumes that the distribution of each anthropometric index in a population is always perfectly normal and that the standard deviation is always exactly one. This assumption is almost always violated. If it is violated then the use of the SMART flagging criteria may lead to records being flagged inappropriately. There are ways (e.g. transforming data toward normality, using the sample standard deviation) to avoid this problem but using them would also not be using the SMART flagging criteria.
Wide-area surveys such as MICS and DHS will usually collect data from many populations. Each population may have different distributions of anthropometric indices and different prevalence of anthropometric indicators. In this case the mean of the entire survey sample will not be a suitable reference mean and the assumed standard deviation (i.e. SD = 1) will usually be too narrow to set limits that define statistical outliers. This will lead to records being flagged incorrectly. This is illustrated in Figure F3. Stratum or district specific means should be used instead of whole sample means, but this may not solve the problem entirely.
If SMART flagging criteria have already been applied to data and the flagged records have been removed from the dataset then a subsequent application of the SMART flagging criteria will tend to flag additional records. SMART flagging criteria should, therefore, only be applied to raw data. Do not apply SMART flagging criteria to data from which flagged records have been removed.
It is important to note that only one set of flagging criteria, either WHO or SMART, should be used at any one time.
The WHO and SMART flagging criteria are designed to be applied to survey samples. They should not be applied to clinical populations or samples.
Software such as ENA from SMART, EpiInfo, WHO Anthro, WHO AnthroPlus, and scripts / macros for R, SAS, SPSS, and STATA provided by the WHO are frequently used to calculate anthropometric indices and apply flagging criteria to data from surveys that collect anthropometric data. It is quite common to receive data to which flagging criteria have already been applied and contain one or more flag variables. You may use these flags if you are sure which flagging criteria have been applied. If you are unsure which flagging criteria have been applied then you should apply your flagging criteria of choice using one of these software packages or the procedures outlined in this section. You may also need to recalculate anthropometric indices using WHO reference values if they were calculated using NCHS, CDC, or local growth references.
Applying WHO flagging criteria to survey data
For a first exercise, we will apply the WHO flagging criteria to survey data.
We will retrieve a survey dataset:
svy <- read.table("flag.ex01.csv", header = TRUE, sep = ",")#> psu child age sex weight height muac oedema haz waz whz
#> 1 1 1 20 2 6.1 82.5 127 2 -0.07 -4.54 -6.03
#> 2 1 2 13 2 6.4 70.4 116 2 -1.83 -3.04 -2.93
#> 3 1 3 15 1 7.1 67.5 124 2 -4.60 -3.34 -1.25
#> 4 1 4 15 1 7.2 75.4 130 2 -1.48 -3.22 -3.57
#> 5 1 5 15 1 7.4 70.0 124 2 -3.61 -2.99 -1.61
#> 6 1 6 18 2 7.7 70.6 130 2 -3.48 -2.40 -0.82The file flag.ex01.csv is a comma-separated-value (CSV) file containing anthropometric data from a recent SMART survey in Sudan.
Applying WHO flagging criteria is straightforward. We first create a column that will contain the flag code and set this to zero (i.e. no flags) for all records:
svy$flag <- 0#> psu child age sex weight height muac oedema haz waz whz flag
#> 1 1 1 20 2 6.1 82.5 127 2 -0.07 -4.54 -6.03 0
#> 2 1 2 13 2 6.4 70.4 116 2 -1.83 -3.04 -2.93 0
#> 3 1 3 15 1 7.1 67.5 124 2 -4.60 -3.34 -1.25 0
#> 4 1 4 15 1 7.2 75.4 130 2 -1.48 -3.22 -3.57 0
#> 5 1 5 15 1 7.4 70.0 124 2 -3.61 -2.99 -1.61 0
#> 6 1 6 18 2 7.7 70.6 130 2 -3.48 -2.40 -0.82 0Then we apply the flagging criteria for each index. Here we apply the WHO flagging criteria to the HAZ index:
#> psu child age sex weight height muac oedema haz waz whz flag
#> 1 1 1 20 2 6.1 82.5 127 2 -0.07 -4.54 -6.03 0
#> 2 1 2 13 2 6.4 70.4 116 2 -1.83 -3.04 -2.93 0
#> 3 1 3 15 1 7.1 67.5 124 2 -4.60 -3.34 -1.25 0
#> 4 1 4 15 1 7.2 75.4 130 2 -1.48 -3.22 -3.57 0
#> 5 1 5 15 1 7.4 70.0 124 2 -3.61 -2.99 -1.61 0
#> 6 1 6 18 2 7.7 70.6 130 2 -3.48 -2.40 -0.82 0This can be translated as “if HAZ is not missing and HAZ is below -6 or HAZ is above +6 then add 1 to the flag variable else leave the flag variable unchanged”.
Be careful when using the \(<\) comparison operator with negative numbers. Always insert a space between the \(<\) and \(–\) characters. R interprets \(<-\) as an assignment operator and may produce unexpected and unwanted results without issuing a warning or error message.
Here we apply the WHO flagging criteria to the WHZ index:
#> psu child age sex weight height muac oedema haz waz whz flag
#> 1 1 1 20 2 6.1 82.5 127 2 -0.07 -4.54 -6.03 2
#> 2 1 2 13 2 6.4 70.4 116 2 -1.83 -3.04 -2.93 0
#> 3 1 3 15 1 7.1 67.5 124 2 -4.60 -3.34 -1.25 0
#> 4 1 4 15 1 7.2 75.4 130 2 -1.48 -3.22 -3.57 0
#> 5 1 5 15 1 7.4 70.0 124 2 -3.61 -2.99 -1.61 0
#> 6 1 6 18 2 7.7 70.6 130 2 -3.48 -2.40 -0.82 0Here we apply the WHO flagging criteria to the WAZ index:
#> psu child age sex weight height muac oedema haz waz whz flag
#> 1 1 1 20 2 6.1 82.5 127 2 -0.07 -4.54 -6.03 2
#> 2 1 2 13 2 6.4 70.4 116 2 -1.83 -3.04 -2.93 0
#> 3 1 3 15 1 7.1 67.5 124 2 -4.60 -3.34 -1.25 0
#> 4 1 4 15 1 7.2 75.4 130 2 -1.48 -3.22 -3.57 0
#> 5 1 5 15 1 7.4 70.0 124 2 -3.61 -2.99 -1.61 0
#> 6 1 6 18 2 7.7 70.6 130 2 -3.48 -2.40 -0.82 0Note that each time we apply a flagging criteria we increase the value of the flagging variable by the next power of two when a problem is detected:
We started with zero
Then we added $2 ^ 0$ (i.e. 1) if HAZ was out of range.
Then we added $2 ^ 1$ (i.e. 2) if WHZ was out of range.
Then we added $2 ^ 2$ (i.e. 4) if WAZ was out of range.If we had another index then we would use \(2 ^ 3\) (i.e. 8) to flag a problem in that index.
The advantage of using this coding scheme is that it compactly codes all possible combinations of problems in a single variable (see table below).
There are a number of flagged records in the example dataset. This:
table(svy$flag)returns:
#>
#> 0 1 2 3 5 6
#> 751 9 12 9 2 3This table shows the relative frequency of detected problems. See table below to find the meaning of each of the codes.
| Code | HAZ | WHZ | WAZ | Suggested action(s) |
|---|---|---|---|---|
| 0 | O | O | O | None |
| 1 | X | O | O | Check height and age |
| 2 | O | X | O | Check weight and height |
| 3 | X | X | O | Check height |
| 4 | O | O | X | Check weight and age |
| 5 | X | O | X | Check age |
| 6 | O | X | X | Check weight |
| 7 | X | X | X | Check age, height, and weight |
The number of flagged records can be found using:
table(svy$flag != 0)["TRUE"]which returns:
#> TRUE
#> 35The proportion of records that are flagged can be found using:
prop.table(table(svy$flag != 0))["TRUE"]This returns:
#> TRUE
#> 0.04452926About 4.45% of records are flagged.
Note that missing values are not flagged. It can be useful to check missing values to see if there are missing component measurements or if a component measurement is out of range for the calculation of index values (e.g. WAZ is only calculated for children aged ten years or younger). This issue can be explored by selection and listing. For example:
This returns:
#> weight height whz
#> 8 8.1 NA NAThere is one missing value for whz in record 8.This is due to a missing value for height (shown as NA). and haz will also be missing. It may be possible to fix this issue if the missing data are available on paper forms.
Flagging has a dual role:
It is a data-checking tool. If you have access to data collection forms you will be often able to check records and fix data-entry errors in the data.
It is a measure of data-quality. Flagged records can indicate problems with measurement, recording, data-entry, and data-checking. The proportion of flagged records in a dataset should, ideally, be below about 2.5%. SMART guidelines consider proportions above 7.5% to be problematic. We found that 4.45% of records in the example dataset were flagged. The data are of acceptable quality.
We can use:
svy[svy$flag != 0, ]
#> psu child age sex weight height muac oedema haz waz whz flag
#> 1 1 1 20 2 6.1 82.5 127 2 -0.07 -4.54 -6.03 2
#> 29 1 29 24 2 16.3 107.3 155 2 6.69 2.69 -0.82 1
#> 32 2 1 12 1 6.1 99.4 112 2 9.95 -4.02 -9.18 3
#> 35 2 4 24 2 6.8 65.5 128 2 -6.27 -4.30 -0.63 1
#> 88 3 30 24 2 16.9 107.5 158 2 6.75 2.95 -0.47 1
#> 106 4 18 36 1 13.4 65.7 152 2 -8.20 -0.56 7.64 3
#> 174 7 3 36 2 6.8 66.6 134 2 -7.47 -5.35 -1.01 1
#> 198 8 1 27 2 5.5 66.0 112 2 -6.59 -5.92 -3.27 1
#> 280 11 7 24 1 6.7 81.7 140 2 -1.77 -4.86 -5.63 2
#> 286 11 13 48 1 9.4 77.3 146 2 -6.21 -4.25 -0.69 1
#> 292 11 19 12 1 12.9 92.3 152 2 6.97 2.68 -0.50 1
#> 307 12 3 36 1 7.5 90.0 130 2 -1.64 -4.99 -6.42 2
#> 350 14 1 20 1 5.7 77.8 142 2 -2.27 -5.49 -6.47 2
#> 352 14 3 48 1 6.5 80.7 140 2 -5.40 -6.22 -5.74 6
#> 368 14 19 48 1 13.4 66.3 144 2 -8.83 -1.58 7.33 3
#> 399 15 21 36 1 14.3 66.0 154 2 -8.12 -0.02 8.58 3
#> 400 15 22 48 1 14.5 68.0 152 2 -8.42 -0.95 7.80 3
#> 405 16 4 24 2 7.8 65.0 145 2 -6.42 -3.27 1.04 1
#> 406 16 5 12 1 7.8 98.0 138 2 9.36 -1.93 -7.23 3
#> 408 16 7 48 1 8.0 77.0 128 2 -6.28 -5.20 -2.66 1
#> 432 17 3 6 1 7.9 98.4 138 2 14.38 -0.04 -7.18 3
#> 433 17 4 48 2 8.3 94.9 136 2 -1.82 -4.79 -5.63 2
#> 490 19 1 12 2 5.3 72.0 152 2 -0.78 -4.27 -5.30 2
#> 591 22 24 36 1 14.0 69.0 152 2 -7.31 -0.20 6.77 3
#> 594 23 1 36 1 5.4 80.0 140 2 -4.34 -6.66 -7.27 6
#> 595 23 2 36 1 5.9 72.0 114 2 -6.50 -6.26 -4.96 5
#> 596 23 3 24 1 6.3 77.0 130 2 -3.31 -5.24 -5.38 2
#> 599 23 6 36 1 6.5 80.0 130 2 -4.34 -5.79 -5.61 2
#> 616 23 23 36 1 16.0 74.0 144 2 -5.96 0.90 6.82 2
#> 640 25 1 12 2 6.3 99.3 110 2 9.82 -2.96 -8.25 3
#> 641 25 2 48 2 6.7 85.0 140 2 -4.12 -5.90 -5.83 2
#> 671 26 1 48 1 5.3 95.0 135 2 -1.99 -7.03 -9.71 6
#> 690 26 20 36 1 16.0 79.0 162 2 -4.61 0.90 5.34 2
#> 715 28 4 36 2 7.7 103.0 114 2 2.09 -4.60 -7.31 2
#> 757 30 1 24 1 5.5 68.6 106 2 -6.06 -6.01 -4.76 5to display the flagged records.
This:
svy[svy$flag != 0, c("psu", "child", "flag")]
#> psu child flag
#> 1 1 1 2
#> 29 1 29 1
#> 32 2 1 3
#> 35 2 4 1
#> 88 3 30 1
#> 106 4 18 3
#> 174 7 3 1
#> 198 8 1 1
#> 280 11 7 2
#> 286 11 13 1
#> 292 11 19 1
#> 307 12 3 2
#> 350 14 1 2
#> 352 14 3 6
#> 368 14 19 3
#> 399 15 21 3
#> 400 15 22 3
#> 405 16 4 1
#> 406 16 5 3
#> 408 16 7 1
#> 432 17 3 3
#> 433 17 4 2
#> 490 19 1 2
#> 591 22 24 3
#> 594 23 1 6
#> 595 23 2 5
#> 596 23 3 2
#> 599 23 6 2
#> 616 23 23 2
#> 640 25 1 3
#> 641 25 2 2
#> 671 26 1 6
#> 690 26 20 2
#> 715 28 4 2
#> 757 30 1 5produces a more compact list.
In the example dataset records are identified using a combination of the psu and child variables.
The listed records can be checked and edited (see previous table). Anthropometric indices can then be recalculated and the flagging process repeated until all records that can be fixed have been fixed.
Records that cannot be fixed can be censored during analysis. Records are usually censored on an index-by-index basis. For example, an analysis based on WHZ would censor records in which the flag variable is 2, 3, 6, or 7.
Table below shows censoring rules for each index:
| Analysis uses … | Censor if flag code is … |
|---|---|
| HAZ | 1, 3, 5, or 7 |
| WHZ | 2, 3, 6, or 7 |
| WAZ | 4, 5, 6, or 7 |
You should be very careful when applying censoring rules. An analysis of prevalence using WHZ, for example, will usually include children with oedema because a commonly used case-definition for acute malnutrition is:
\[ \text{WHZ} < -2 ~ \text{or bilateral pitting oedema} \]
If you want to use case-definitions that include oedema then you should be careful not to exclude children with oedema when censoring flagged records. For an analysis using WAZ you might want to exclude oedema cases.
Applying SMART flagging criteria to survey data
In the next exercise we will apply SMART flagging criteria to the same survey dataset.
We will retrieve the survey dataset:
svy <- read.table("flag.ex01.csv", header = TRUE, sep = ",")#> psu child age sex weight height muac oedema haz waz whz
#> 1 1 1 20 2 6.1 82.5 127 2 -0.07 -4.54 -6.03
#> 2 1 2 13 2 6.4 70.4 116 2 -1.83 -3.04 -2.93
#> 3 1 3 15 1 7.1 67.5 124 2 -4.60 -3.34 -1.25
#> 4 1 4 15 1 7.2 75.4 130 2 -1.48 -3.22 -3.57
#> 5 1 5 15 1 7.4 70.0 124 2 -3.61 -2.99 -1.61
#> 6 1 6 18 2 7.7 70.6 130 2 -3.48 -2.40 -0.82and create a column that will contain the flag code and set this to zero (i.e. no flags) for all records:
svy$flag <- 0Applying SMART flagging criteria requires us to first calculate a mean index value:
meanHAZ <- mean(svy$haz, na.rm = TRUE)and then to use this mean value to define flagging ranges:
svy$flag <- ifelse(!is.na(svy$haz) &
(svy$haz < (meanHAZ - 3) | svy$haz > (meanHAZ + 3)),
svy$flag + 1, svy$flag)We do this for each index:
meanWHZ <- mean(svy$whz, na.rm = TRUE)
svy$flag <- ifelse(!is.na(svy$whz) &
(svy$whz < (meanWHZ - 3) | svy$whz > (meanWHZ + 3)),
svy$flag + 2, svy$flag)
meanWAZ <- mean(svy$waz, na.rm = TRUE)
svy$flag <- ifelse(!is.na(svy$waz) &
(svy$waz < (meanWAZ - 3) | svy$waz > (meanWAZ + 3)),
svy$flag + 4, svy$flag)There are a number of flagged records in the example dataset.
This:
table(svy$flag)returns:
#>
#> 0 1 2 3 4 5 6 7
#> 660 59 11 16 1 19 16 4This table shows the relative frequency of detected problems. See the previous table to find the meaning of each of the codes. The number of flagged records can be found using:
table(svy$flag != 0)["TRUE"]which returns:
#> TRUE
#> 126The proportion of records that are flagged can be found using:
prop.table(table(svy$flag != 0))["TRUE"]which returns:
#> TRUE
#> 0.1603053About 16% of records are flagged. This is a very high proportion of records flagged.
Note how the SMART flagging criteria identify considerably more records (126 records flagged) than the WHO flagging criteria (35 records flagged). In this example the SMART flagging criteria flagged 91 biologically plausible records.
We can list flagged records using:
svy[svy$flag != 0, ]
#> psu child age sex weight height muac oedema haz waz whz flag
#> 1 1 1 20 2 6.1 82.5 127 2 -0.07 -4.54 -6.03 2
#> 3 1 3 15 1 7.1 67.5 124 2 -4.60 -3.34 -1.25 1
#> 15 1 15 36 1 12.3 79.7 144 2 -4.42 -1.27 1.97 3
#> 28 1 28 48 2 15.8 109.7 146 2 1.62 -0.12 -1.72 1
#> 29 1 29 24 2 16.3 107.3 155 2 6.69 2.69 -0.82 5
#> 31 1 31 48 2 18.8 109.9 166 2 1.66 1.10 0.13 1
#> 32 2 1 12 1 6.1 99.4 112 2 9.95 -4.02 -9.18 3
#> 34 2 3 24 2 6.5 76.0 108 2 -3.01 -4.61 -4.16 6
#> 35 2 4 24 2 6.8 65.5 128 2 -6.27 -4.30 -0.63 1
#> 36 2 5 36 1 7.3 76.0 110 2 -5.42 -5.15 -3.56 5
#> 42 2 11 12 2 9.9 80.0 150 2 2.32 0.82 -0.21 1
#> 44 2 13 36 2 10.5 78.0 142 2 -4.48 -2.24 0.87 1
#> 52 2 21 36 1 12.7 77.5 144 2 -5.01 -1.01 2.77 3
#> 57 2 26 24 1 15.5 93.7 166 2 2.16 2.13 1.46 5
#> 59 3 1 18 2 5.7 67.0 110 2 -4.72 -4.72 -3.21 5
#> 66 3 8 48 2 9.4 79.0 144 2 -5.51 -4.03 -0.57 1
#> 76 3 18 24 2 12.1 96.0 138 2 3.19 0.42 -1.79 1
#> 88 3 30 24 2 16.9 107.5 158 2 6.75 2.95 -0.47 5
#> 89 4 1 26 2 6.6 71.7 114 2 -4.73 -4.74 -2.95 5
#> 106 4 18 36 1 13.4 65.7 152 2 -8.20 -0.56 7.64 3
#> 107 4 19 24 1 13.7 97.6 150 2 3.43 1.05 -0.89 1
#> 122 5 4 24 1 8.0 73.3 130 2 -4.52 -3.61 -1.66 1
#> 125 5 7 36 2 11.3 106.2 150 2 2.93 -1.63 -4.61 3
#> 139 5 21 24 2 15.2 82.0 138 2 -1.15 2.18 3.97 6
#> 154 6 14 24 2 11.9 91.0 148 2 1.64 0.29 -0.91 1
#> 165 6 25 36 1 14.9 108.0 144 2 3.21 0.31 -2.13 1
#> 173 7 2 10 2 6.5 76.2 122 2 1.91 -2.23 -4.20 3
#> 174 7 3 36 2 6.8 66.6 134 2 -7.47 -5.35 -1.01 5
#> 187 7 16 10 2 11.6 84.3 152 2 5.19 2.50 0.54 5
#> 198 8 1 27 2 5.5 66.0 112 2 -6.59 -5.92 -3.27 5
#> 199 8 2 24 2 6.4 75.0 138 2 -3.32 -4.72 -4.10 6
#> 201 8 4 24 1 7.1 70.5 122 2 -5.44 -4.47 -2.31 1
#> 203 8 6 31 1 8.5 72.9 134 2 -5.71 -3.83 -0.79 1
#> 205 8 8 36 2 9.4 78.0 146 2 -4.48 -3.18 -0.35 1
#> 212 8 15 48 1 11.4 102.5 126 2 -0.20 -2.88 -4.22 2
#> 254 9 30 42 1 17.9 109.4 164 2 2.41 1.23 -0.26 1
#> 255 10 1 23 1 6.7 71.0 118 2 -5.32 -4.76 -3.23 5
#> 274 11 1 24 2 5.8 71.9 108 2 -4.28 -5.34 -4.40 6
#> 280 11 7 24 1 6.7 81.7 140 2 -1.77 -4.86 -5.63 6
#> 283 11 10 36 1 8.5 78.3 126 2 -4.80 -4.20 -2.19 1
#> 286 11 13 48 1 9.4 77.3 146 2 -6.21 -4.25 -0.69 1
#> 290 11 17 24 2 12.4 99.9 136 2 4.40 0.62 -2.33 1
#> 292 11 19 12 1 12.9 92.3 152 2 6.97 2.68 -0.50 5
#> 301 11 28 24 2 15.1 85.3 140 2 -0.13 2.13 2.94 6
#> 302 11 29 30 1 15.2 82.9 154 2 -2.65 1.13 3.76 2
#> 303 11 30 48 2 15.8 90.5 132 2 -2.84 -0.12 2.29 2
#> 307 12 3 36 1 7.5 90.0 130 2 -1.64 -4.99 -6.42 6
#> 313 12 9 12 1 10.0 81.0 150 2 2.21 0.33 -0.75 1
#> 315 12 11 48 1 10.6 84.0 142 2 -4.61 -3.43 -0.75 1
#> 330 13 3 24 1 7.7 73.0 114 2 -4.62 -3.90 -2.06 1
#> 340 13 13 12 2 11.1 79.0 152 2 1.94 1.72 1.26 5
#> 345 13 18 24 1 13.3 96.1 142 2 2.94 0.79 -0.93 1
#> 350 14 1 20 1 5.7 77.8 142 2 -2.27 -5.49 -6.47 6
#> 352 14 3 48 1 6.5 80.7 140 2 -5.40 -6.22 -5.74 7
#> 366 14 17 24 1 12.7 92.3 185 2 1.70 0.39 -0.70 1
#> 368 14 19 48 1 13.4 66.3 144 2 -8.83 -1.58 7.33 3
#> 379 15 1 12 1 5.1 66.0 106 2 -4.10 -5.24 -4.75 6
#> 395 15 17 24 1 13.1 80.0 144 2 -2.33 0.66 2.62 2
#> 399 15 21 36 1 14.3 66.0 154 2 -8.12 -0.02 8.58 3
#> 400 15 22 48 1 14.5 68.0 152 2 -8.42 -0.95 7.80 3
#> 403 16 2 24 1 7.0 74.0 130 2 -4.29 -4.57 -3.56 4
#> 405 16 4 24 2 7.8 65.0 145 2 -6.42 -3.27 1.04 1
#> 406 16 5 12 1 7.8 98.0 138 2 9.36 -1.93 -7.23 3
#> 408 16 7 48 1 8.0 77.0 128 2 -6.28 -5.20 -2.66 5
#> 432 17 3 6 1 7.9 98.4 138 2 14.38 -0.04 -7.18 3
#> 433 17 4 48 2 8.3 94.9 136 2 -1.82 -4.79 -5.63 6
#> 435 17 6 9 1 8.8 77.7 136 2 2.55 -0.11 -1.61 1
#> 448 17 19 36 1 13.9 105.0 138 2 2.41 -0.26 -2.34 1
#> 449 17 20 36 2 14.4 107.5 162 2 3.27 0.30 -2.27 1
#> 460 17 31 48 1 18.5 96.2 170 2 -1.70 0.96 3.00 2
#> 462 18 2 7 1 7.6 76.5 146 2 3.38 -0.80 -3.19 1
#> 464 18 4 23 1 8.0 73.4 134 2 -4.52 -3.49 -1.69 1
#> 468 18 8 36 1 9.3 77.6 140 2 -4.99 -3.57 -0.89 1
#> 483 18 23 24 1 15.8 102.5 146 2 5.04 2.29 -0.21 5
#> 489 18 29 48 2 19.2 109.9 164 2 1.66 1.24 0.35 1
#> 490 19 1 12 2 5.3 72.0 152 2 -0.78 -4.27 -5.30 2
#> 499 19 10 48 1 10.0 84.2 140 2 -4.56 -3.84 -1.53 1
#> 508 19 19 24 1 13.7 98.0 180 2 3.56 1.05 -0.97 1
#> 510 19 21 24 1 13.9 92.7 152 2 1.83 1.18 0.35 1
#> 512 19 23 36 2 15.8 101.5 174 2 1.69 1.00 0.09 1
#> 519 20 7 18 1 9.4 69.5 140 2 -4.73 -1.36 1.47 1
#> 528 20 16 24 2 12.5 91.5 146 2 1.79 0.68 -0.46 1
#> 530 20 18 24 2 13.2 91.2 160 2 1.70 1.11 0.22 1
#> 536 20 24 48 2 17.5 109.9 154 2 1.66 0.61 -0.63 1
#> 537 20 25 36 1 18.1 109.3 162 2 3.57 1.90 -0.11 5
#> 557 21 19 24 2 11.4 92.0 138 2 1.95 -0.05 -1.64 1
#> 587 22 20 36 2 12.7 80.4 154 2 -3.85 -0.68 2.38 2
#> 591 22 24 36 1 14.0 69.0 152 2 -7.31 -0.20 6.77 3
#> 594 23 1 36 1 5.4 80.0 140 2 -4.34 -6.66 -7.27 6
#> 595 23 2 36 1 5.9 72.0 114 2 -6.50 -6.26 -4.96 7
#> 596 23 3 24 1 6.3 77.0 130 2 -3.31 -5.24 -5.38 6
#> 598 23 5 24 2 6.5 71.0 124 2 -4.56 -4.61 -2.93 5
#> 599 23 6 36 1 6.5 80.0 130 2 -4.34 -5.79 -5.61 6
#> 600 23 7 24 2 7.0 70.0 112 2 -4.87 -4.10 -1.75 1
#> 604 23 11 14 1 8.0 66.0 136 2 -4.86 -2.11 0.77 1
#> 607 23 14 36 1 8.3 74.0 138 2 -5.96 -4.36 -1.40 1
#> 612 23 19 48 1 11.5 80.0 144 2 -5.56 -2.81 1.14 1
#> 616 23 23 36 1 16.0 74.0 144 2 -5.96 0.90 6.82 3
#> 621 24 5 24 1 8.4 72.2 140 2 -4.88 -3.22 -0.73 1
#> 633 24 17 24 2 12.9 93.2 152 2 2.32 0.93 -0.46 1
#> 640 25 1 12 2 6.3 99.3 110 2 9.82 -2.96 -8.25 3
#> 641 25 2 48 2 6.7 85.0 140 2 -4.12 -5.90 -5.83 6
#> 649 25 10 36 2 8.6 78.0 134 2 -4.48 -3.85 -1.38 1
#> 661 25 22 24 2 12.4 91.0 140 2 1.64 0.62 -0.44 1
#> 671 26 1 48 1 5.3 95.0 135 2 -1.99 -7.03 -9.71 6
#> 672 26 2 18 2 5.6 67.0 108 2 -4.72 -4.84 -3.41 5
#> 674 26 4 36 1 8.0 76.0 134 2 -5.42 -4.60 -2.40 5
#> 679 26 9 48 1 10.5 82.0 142 2 -5.09 -3.50 -0.38 1
#> 683 26 13 24 1 13.8 75.0 156 2 -3.97 1.11 4.39 2
#> 685 26 15 24 1 14.3 85.0 168 2 -0.69 1.42 2.40 2
#> 689 26 19 36 1 15.8 104.0 148 2 2.14 0.80 -0.54 1
#> 690 26 20 36 1 16.0 79.0 162 2 -4.61 0.90 5.34 3
#> 692 27 2 24 2 7.1 68.2 124 2 -5.43 -4.00 -1.04 1
#> 698 27 8 36 2 8.4 75.4 124 2 -5.16 -4.01 -1.06 1
#> 715 28 4 36 2 7.7 103.0 114 2 2.09 -4.60 -7.31 7
#> 721 28 10 48 1 10.3 82.0 148 2 -5.09 -3.64 -0.61 1
#> 723 28 12 15 1 11.0 73.0 162 2 -2.43 0.59 2.24 2
#> 733 29 1 16 1 5.9 69.2 112 2 -4.26 -4.85 -4.17 6
#> 734 29 2 17 1 6.1 69.3 114 2 -4.53 -4.75 -3.81 5
#> 745 29 13 24 1 11.0 70.3 114 2 -5.50 -0.87 3.01 3
#> 757 30 1 24 1 5.5 68.6 106 2 -6.06 -6.01 -4.76 7
#> 767 30 11 36 2 10.2 77.5 142 2 -4.61 -2.49 0.66 1
#> 781 30 25 24 2 13.3 91.5 152 2 1.79 1.16 0.24 1
#> 783 30 27 36 1 14.2 102.3 138 2 1.68 -0.08 -1.48 1
#> 784 30 28 36 1 14.6 106.1 154 2 2.70 0.15 -1.97 1
#> 786 30 30 36 2 15.5 101.2 154 2 1.61 0.86 -0.05 1The listed records can be checked and edited (see previous table). Anthropometric indices can then be recalculated and the flagging process repeated until all records that can be fixed have been fixed.
When listing records or displaying very large tables you may see a message like this:
#> [1] "[ reached getOption(\"max.print\") -- omitted 43 rows ]"The max.print option sets a limit on the length of information that can be displayed by a single command. You can alter this behaviour using:
options(max.print = 99999)Flagging data from older children
The process of flagging anthropometric indices in older children is very similar to that used with younger children.
We will retrieve a survey dataset:
svy <- read.table("flag.ex02.csv", header = TRUE, sep = ",") #> school sex ageMonths weight height haz baz
#> 1 1112 1 173 25.5 179.0 1.70 -8.19
#> 2 1113 2 145 22.7 164.0 1.79 -6.81
#> 3 1116 1 150 13.5 135.0 -2.40 -8.64
#> 4 1123 1 150 25.3 165.0 1.73 -6.92
#> 5 1404 2 163 19.0 116.5 -6.05 -2.89
#> 6 1501 2 185 27.4 136.6 -3.73 -2.85The file flag.ex02.csv is a comma-separated-value (CSV) file containing anthropometric data from a survey of children aged 11 year or older and attending school in Ethiopia.
The variables of interest are height-for-age z-score (haz) and BMI-for-age z-score (baz). We will apply the WHO flagging criteria (see previous table) to these variables:
svy$flag <- 0
svy$flag <- ifelse(!is.na(svy$haz) & (svy$haz < -6 | svy$haz > 6),
svy$flag + 1, svy$flag)
svy$flag <- ifelse(!is.na(svy$baz) & (svy$baz < -5 | svy$baz > 5),
svy$flag + 2, svy$flag)Note that we do not usually apply SMART flagging criteria to older (i.e. > 59 months) children.
The coding of the flag variable is shown in previous table.
| Code | HAZ | BAZ | Suggested action(s) |
|---|---|---|---|
| 0 | O | O | None |
| 1 | X | O | Check height and age |
| 2 | O | X | Check weight, height and age |
| 3 | X | X | Check weight, height and age |
This:
table(svy$flag)returns:
#>
#> 0 1 2
#> 960 2 11This table shows the relative frequency of detected problems. See previous table to find the meaning of each of the codes. The number of flagged records can be found using:
table(svy$flag != 0)["TRUE"]which returns:
#> TRUE
#> 13The proportion of records that are flagged can be found using:
prop.table(table(svy$flag != 0))["TRUE"]which returns:
#> TRUE
#> 0.01336074About 1.3% of records are flagged. This is an acceptably low proportion of records flagged. We can list flagged records using:
svy[svy$flag != 0, ]
#> school sex ageMonths weight height haz baz flag
#> 1 1112 1 173 25.5 179.0 1.70 -8.19 2
#> 2 1113 2 145 22.7 164.0 1.79 -6.81 2
#> 3 1116 1 150 13.5 135.0 -2.40 -8.64 2
#> 4 1123 1 150 25.3 165.0 1.73 -6.92 2
#> 5 1404 2 163 19.0 116.5 -6.05 -2.89 1
#> 23 1501 2 137 24.7 155.0 1.09 -5.20 2
#> 190 1507 1 173 24.0 154.0 -1.52 -6.46 2
#> 328 1511 1 138 26.9 165.5 2.82 -6.29 2
#> 969 1705 1 185 27.4 150.4 -2.62 -5.06 2
#> 970 1708 1 197 23.9 126.2 -6.19 -3.17 1
#> 971 1708 1 185 23.6 140.7 -3.86 -5.21 2
#> 972 1909 2 174 26.5 153.7 -1.04 -5.04 2
#> 973 2001 1 139 20.7 143.1 -0.49 -6.02 2The listed records can be checked and edited (see previous table). Anthropometric indices can then be recalculated and the flagging process repeated until all records that can be fixed have been fixed.