Precision indicators (standard deviation, coefficient of variation, 95% confidence interval) take into account errors due to sampling, correction of total non-response, calibration and winsorizing, but not errors related to the correction of partial non-responses.

82 variables have been defined for precision calculations in ICT-VSE 2016:
- 71 qualitative variables
- yes/no qualitative variables
- 2 quantitative variables as a percentage transformed into "qualitative" variables:
- the share of enterprises with BTOC>0
- the share of enterprises with  BTOB>0
- 4 quantitative variables calculated in the form of a ratio:
- the share of persons employed using a computer
- the share of persons employed using a computer with Internet access
- the share of web sales in the T/O (VENT_WEB_VAL / CA)
- the share of web B to C sales in T/O (VENT_WEB_VAL x BTOC / CA)
- 5 quantitative variables in volume:
- CA
- EFF
- EMP_ORDI_VAL
- EMP_INTERNET_VAL
- VENT_WEB_VAL.

For qualitative variables, the indicators are:
average (%) = proportion of "yes" relative to all the units (concerned or not)
standard deviation(s) = root(variance)
95% confidence interval (CI) = average +/- 1.96 s.

For quantitative variables transformed into qualitative, the indicators are:
average (%) = proportion of enterprises having BTOC/BTOB >0, relative to all the units (concerned or not)
standard deviation(s) = root(variance)
95% confidence interval (CI) = average +/- 1.96 s.

For quantitative variables in the form of a ratio, the indicators are:
ratio (%) = amount of web sales/web B to C sales relative to T/O of all the units (concerned or not)
standard deviation(s) = root(variance)
95% confidence interval (CI) = average +/- 1.96 s.

For quantitative variables in volume, the indicators are:
total coefficient of variation (CV) = s / total
95% confidence interval (CI)  = total +/- 1.96 s.

Note: For quantitative variables in volume, the lower limit of the confidence interval is sometimes negative. This means that the normal approximation used in the formula for calculating the confidence interval (the factor 1.96) is not valid. In this case, the unweighted total observed for the sample was taken as the value for the lower limit, and therefore the confidence interval is no longer centred around the value of the estimator.

The results of the calculations are in file "S.13.1. Calculs_précision_TIC-TPE16.xls"

Cf. files << Écart sur l'agrégat venant du redressement / Ensemble des entreprises >> and << Écart sur l'agrégat venant du redressement / Microentreprises >>, which summarize the differences in aggregate from the correction (including cleaning) for all qualitative or quantitative variables.
The first file deals only with "microenterprises", while the second covers all 5,800 exploitable units.