Standard output formats

Use this guide to format DataLab outputs for clearance. A consistent format helps the clearance team assess requests and check that outputs meet requirements. The long format suits most outputs.

This guide will help you:

format outputs consistently so the clearance team can assess them efficiently
include the evidence needed to show that you have applied confidentiality protections
reduce the chance that we return your request for revision.

It covers common output types, including counts, proportions, ratios, means, quantiles and models, and includes examples.

Format tabular output

Use this format for counts, proportions, means, quantiles and ratios.
See the examples to check how to format your table.
Refer to the reference tables for detailed requirements: Table 1 for output data, Table 2 for variables and Table 3 for evidence.

Step 1: Include the required columns

Create a column for:

each output data type listed in Table 1
each grouping variable listed in Table 2
each required evidence column listed in Tables 2 and 3.

Step 2: Name evidence-only columns

Start the name of each evidence-only column with EVIDENCE_.
The EVIDENCE_ columns will be used to assess your request and will not be cleared.
For example, EVIDENCE_COUNT can show the raw counts used to calculate other output data.

Step 3: Apply suppression where required

Enter SUPPRESSED in each suppressed cell.
Do not suppress values in EVIDENCE_ columns. For suppressed counts, create COUNT_SUPPRESSED and enter SUPPRESSED where required.

Step 4: Save the file as a CSV

Submit your table as a CSV file.

Format model output

Use this format for linear, logistic and other models.
Use one row for each model term, such as an independent variable.
See the examples to check how to format your output.
Refer to Table 4 of the reference tables for detailed requirements.

Step 1: Include the required columns

Create all required columns listed in Table 4 below.

Step 2: Include additional columns where required

For ordinary least squares (OLS) models, include the R2 and IS_CONTINUOUS_VARIABLE columns.
For other statistics, use STAT_[statistic] columns, such as STAT_PVALUE or STAT_SE.

Step 3: Name evidence-only columns

Start the name of each evidence-only column with EVIDENCE_.
The EVIDENCE_ columns will be used to assess your request and will not be cleared.
For example, EVIDENCE_DEGREES_OF_FREEDOM can show the unrounded degrees of freedom used in the model.

Step 4: Apply suppression where required

Enter SUPPRESSED for each suppressed value in the VALUE or STAT_[statistic] column.
Do not suppress values in EVIDENCE_ columns.

Step 5: Save the file as a CSV

Submit your output as a CSV file.

Reference tables

Use these reference tables to create your output and apply the required column definitions and validation rules.

Reference for tabular outputs

Table 1: Data columns

Column name	Description	Requirements
COUNT	The unweighted, unrounded count of units, such as people, businesses or households, in the intersection of all VAR_[variable] columns	Rule of 10 May require COUNT_SECONDARY_CONTRIBUTORS
COUNT_SUPPRESSED	COUNT after applying the rule of 10, including consequential suppression	Requires COUNT COUNT must be prefixed with EVIDENCE_.
COUNT_WEIGHTED	Weighted COUNT	Requires COUNT
COUNT_ROUNDED	Rounded COUNT Must be rounded using one of these methods: Random 5 (for example, 16 rounds to 15 or 20 with equal chance) Nearest 10 (for example, 16 rounds to 20) Nearest multiple of greater than 10 (for example, round to the nearest 100)	Requires COUNT COUNT must be prefixed with EVIDENCE_.
PROPORTION	COUNT divided by COUNT_DENOMINATOR COUNT must be a subset of COUNT_DENOMINATOR	Requires COUNT, COUNT_DENOMINATOR and COUNT_COMPLEMENT Must be in decimal format (0.xx), not a percentage (xx%)
PROPORTION_WEIGHTED	COUNT_WEIGHTED divided by COUNT_DENOMINATOR_WEIGHTED	Requires COUNT_WEIGHTED, COUNT_DENOMINATOR_WEIGHTED
PROPORTION_ROUNDED	COUNT_ROUNDED divided by COUNT_DENOMINATOR_ROUNDED	Requires COUNT_ROUNDED and COUNT_DENOMINATOR_ROUNDED
RATIO	COUNT divided by COUNT_DENOMINATOR COUNT does not need to be a subset of COUNT_DENOMINATOR	Requires COUNT and COUNT_DENOMINATOR Must be in decimal format (x.xx), not a percentage (xxx%)
RATIO_WEIGHTED	COUNT_WEIGHTED divided by COUNT_DENOMINATOR_WEIGHTED	Requires COUNT_WEIGHTED and COUNT_DENOMINATOR_WEIGHTED
RATIO_ROUNDED	COUNT_ROUNDED divided by COUNT_DENOMINATOR_ROUNDED	Requires COUNT_ROUNDED and COUNT_DENOMINATOR_ROUNDED
MEAN	The average of OUTCOME_VARIABLE for the intersection of all VAR_[variable] columns	Requires OUTCOME_VARIABLE, COUNT, LARGEST_VALUE, SECOND_LARGEST_VALUE, LARGEST_CONTRIBUTION and TWO_LARGEST_CONTRIBUTIONS If OUTCOME_VARIABLE can have negative values, SUM_ABSOLUTE is required. Otherwise, SUM is required.
SUM	The sum of OUTCOME_VARIABLE for the intersection of all VAR_[variable] columns	Requires OUTCOME_VARIABLE, COUNT, LARGEST_VALUE, SECOND_LARGEST_VALUE, LARGEST_CONTRIBUTION and TWO_LARGEST_CONTRIBUTIONS If OUTCOME_VARIABLE can have negative values, SUM_ABSOLUTE is required.
MEDIAN	The median of OUTCOME_VARIABLE for the intersection of all VAR_[variable] columns This is the same as QUANTILE_0.5.	Requires OUTCOME_VARIABLE and COUNT
QUANTILE_[quantile]	A quantile of OUTCOME_VARIABLE for the intersection of all VAR_[variable] columns For example, QUANTILE_0.25 is the first quartile and QUANTILE_0.99 is the 99th percentile.	Quantile rules Requires OUTCOME_VARIABLE and COUNT

Table 2: Variable columns

Column name	Description	Requirements
VAR_[variable]	Used to group data For example, VAR_AGEGROUP, VAR_BUSINESS_SIZE, VAR_SEIFA_QUINTILE There is no limit to the number of VAR_ columns. For a column named VAR_AGEGROUP, rows may be 0-17, 18-29, 30-64, and 65+.	Must start with VAR_ Can include the following characters: All letters a-z, A-Z All numbers 0-9 Spaces, _, - Totals for a group are indicated by TOTAL in the row. See examples.
VAR_DENOMINATOR_[variable]	Like VAR_[variable] but used to group data for COUNT_DENOMINATOR only There is no limit to the number of VAR_DENOMINATOR_ columns. For a column named VAR_DENOMINATOR_AGEGROUP, rows may be 0-17, 18-29, 30-64, and 65+.	Must start with VAR_DENOMINATOR_ Can include the same characters as VAR_[variable] If requesting PROPORTION, all [variable] of VAR_DENOMINATOR_[variable] must be a subset of VAR_[variable]. For example, If COUNT is grouped by columns VAR_AGEGROUP and VAR_SEX, then COUNT_DENOMINATOR may be grouped by VAR_DENOMINATOR_AGEGROUP, VAR_DENOMINATOR_SEX, or both VAR_DENOMINATOR_AGEGROUP and VAR_DENOMINATOR _SEX. It cannot be grouped by a different VAR_DENOMINATOR_ column. If requesting RATIO, VAR_DENOMINATOR_[variable] can be any variable. It does not need to be a subset of VAR_[variable].
OUTCOME_VARIABLE	The variable used to calculate MEAN, SUM or QUANTILE_[quantile]	COUNT should be the same for all OUTCOME_VARIABLE given the same VAR_[variable]

Table 3: Evidence columns

Column name	Description	Requirements
COUNT_SECONDARY_CONTRIBUTORS	Where relevant, the count of secondary contributors, such as households, employers or registered training organisations	Data-specific rules for secondary contributors Must start with EVIDENCE_
COUNT_DENOMINATOR	The count in the intersection of all VAR_DENOMINATOR_[variable] columns	Rule of 10 Requires VAR_DENOMINATOR_[variable]
COUNT_DENOMINATOR_WEIGHTED	The weighted COUNT_DENOMINATOR	Requires COUNT_DENOMINATOR
COUNT_DENOMINATOR_ROUNDED	Rounded COUNT_DENOMINATOR Must be rounded using one of these methods: Random 5 (for example, 16 rounds to 15 or 20 with equal chance) Nearest 10 (for example, 16 rounds to 20) Nearest multiple of greater than 10 (for example, round to the nearest 100)	Requires COUNT_DENOMINATOR COUNT_DENOMINATOR must be prefixed with EVIDENCE_.
COUNT_COMPLEMENT	COUNT_DENOMINATOR minus COUNT	Rule of 10
SUM_ABSOLUTE	The sum of the absolute values of OUTCOME_VARIABLE for the intersection of all VAR_[variable] columns If all values are positive, this equals SUM.	No additional requirements
LARGEST_VALUE	The largest absolute value of OUTCOME_VARIABLE for the intersection of all VAR_[variable] columns	Must start with EVIDENCE_
SECOND_LARGEST_VALUE	The second-largest absolute value of OUTCOME_VARIABLE for the intersection of all VAR_[variable] columns	Must start with EVIDENCE_
LARGEST_CONTRIBUTION	Equal to LARGEST_VALUE/TOTAL_ABSOLUTE If LARGEST_CONTRIBUTION is less than 0.33 for all rows, then TWO_LARGEST_CONTRIBUTIONS is not required.	Dominance rules: (1,50) rule Must start with EVIDENCE_
TWO_LARGEST_CONTRIBUTIONS	LARGEST_VALUE plus SECOND_LARGEST_VALUE, divided by TOTAL_ABSOLUTE	Dominance rules: (2,67) rule Must start with EVIDENCE_

Reference for model outputs

Table 4: Model output columns

Column name	Description	Requirements
MODEL_NUMBER	The model number Choose a number to identify each model in the output.	Required Must be an integer
MODEL_TYPE	The model type, such as logistic regression or fixed effects model For an ordinary least squares model, enter OLS.	Required If OLS, columns R2 and IS_CONTINUOUS_VARIABLE are required.
DEPENDENT_VARIABLE	The model dependent variable	Required
DEGREES_OF_FREEDOM	The residual degrees of freedom of the model Calculate this as the number of observations minus the number of parameters.	Required Model rules: degrees of freedom
DEGREES_OF_FREEDOM_ROUNDED	Rounded DEGREES_OF_FREEDOM The degrees of freedom may be rounded to prevent differencing. Must be rounded using one of these methods: Random 5 (for example, 16 rounds to 15 or 20 with equal chance) Nearest 10 (for example, 16 rounds to 20) Nearest multiple of greater than 10 (for example, round to the nearest 100)	DEGREES_OF_FREEDOM must be prefixed with EVIDENCE_.
R2	The model R-squared value	Model rules: An OLS model must have an R-squared less than or equal to 0.9.
INDEPENDENT_VARIABLE	A model independent variable	Required
IS_CONTINUOUS_VARIABLE	Enter 1 if the independent variable is continuous. Otherwise, enter 0.	Must be binary
VALUE	The value, estimate or coefficient of the independent variable	Required Requires MODEL_TYPE, MODEL_NUMBER, DEPENDENT_VARIABLE, DEGREES_OF_FREEDOM and INDEPENDENT_VARIABLE
STAT_[statistic]	Other statistics relevant to the model For example, STAT_PVALUE, STAT_STANDARD_ERROR and STAT_CHI2

Data downloads

Standard output format examples (not real data)

Download all (4.94 KB)

Data files

Example 1. Counts
Download csv [223 bytes]
Example 2. Proportions
Download csv [1.07 KB]
Example 3. Means
Download csv [2.22 KB]
Example 4. Quantiles and sums
Download csv [1.16 KB]
Example 5. Means and proportions
Download csv [2.67 KB]
Example 6. Regressions
Download csv [1.42 KB]

APA

Citation

Standard output formats

APA

Citation

Format tabular output

Step 1: Include the required columns

Step 2: Name evidence-only columns

Step 3: Apply suppression where required

Step 4: Save the file as a CSV

Format model output

Step 1: Include the required columns

Step 2: Include additional columns where required

Step 3: Name evidence-only columns

Step 4: Apply suppression where required

Step 5: Save the file as a CSV

Reference tables

Reference for tabular outputs

Table 1: Data columns

Table 2: Variable columns

Table 3: Evidence columns

Reference for model outputs

Table 4: Model output columns

Data downloads

Standard output format examples (not real data)

Example 1. Counts

Example 2. Proportions

Example 3. Means

Example 4. Quantiles and sums

Example 5. Means and proportions

Example 6. Regressions

Provide feedback