Table 2.

Precision and accuracy from the initial standardisation session in Nairobi, Kenya: length data

	TEM*	F test		Sign test***	Bias			Bias
		Lead anthropometrist**	Overall mean**		Lead anthropometrist****	F test*****	Sign test******	Overall mean****	F test*****	Sign test******
Group 1
WHO lead anthropometrist	0.371	–	P > 0.25	P > 0.05	–	–	–	−0.069	P > 0.25	P > 0.05
Observer 1	0.596	0.01 < P < 0.05	0.01 < P < 0.05	P > 0.05	−0.348	0.01 < P < 0.05	P < 0.05	−0.416	0.01 < P < 0.05	P < 0.05
Observer 2	0.314	P > 0.25	P > 0.25	P < 0.05	−0.055	P > 0.25	P > 0.05	−0.124	P > 0.25	P < 0.05
Observer 3	0.368	P > 0.25	P > 0.25	P < 0.05	0.288	P > 0.25	P < 0.05	0.219	P > 0.25	P < 0.05
Observer 4	0.312	P > 0.25	P > 0.25	P < 0.05	0.166	P > 0.25	P < 0.05	0.067	P > 0.25	P < 0.05
Observer 5	0.349	P > 0.25	P > 0.25	P > 0.05	0.432	0.05 < P < 0.10	P < 0.05	0.355	0.10 < P < 0.25	P < 0.05
Group 2
WHO lead anthropometrist	0.311	–	P > 0.25	P > 0.05	–	–	–	0.160	P > 0.25	P < 0.05
Observer 1	0.377	0.10 < P < 0.25	P > 0.25	P > 0.05	−0.288	0.10 < P < 0.25	P < 0.05	−0.127	P > 0.25	P > 0.05
Observer 2	0.534	0.01 < P < 0.05	P > 0.25	P < 0.05	−0.186	P > 0.25	P < 0.05	−0.015	P > 0.25	P > 0.05
Observer 3	0.513	0.01 < P < 0.05	P > 0.25	P > 0.05	−0.227	P > 0.25	P < 0.05	−0.067	P > 0.25	P > 0.05
Observer 4	0.543	P < 0.01	0.10 < P < 0.25	P > 0.05	−0.080	0.05 < P < 0.10	P > 0.05	0.080	0.05 < P < 0.10	P > 0.05
Observer 5	0.409	0.10 < P < 0.25	P > 0.25	P > 0.05	−0.193	P > 0.25	P < 0.05	−0.032	P > 0.25	P > 0.05

^*Technical error of measurement (TEM): $\surd \mathrm{\sum }(d_i^2/2n)$; where d_i is the difference between the i^th participants test and retest measurements by the observer and n is the number of measured participants.

^**F ratio for precision: $\text{Observer}\text{'}\mathrm{s}\mathrm{\sum }(d_i^2)/\text{Lead}\text{anthropometrists}\text{'}\mathrm{s}\mathrm{\sum }(d_i^2)$. When overall mean is the gold standard, d_i in the denominator is the difference between the i^th participant’s overall mean of retest measurements.

^***Precision sign test: binomial proportion p, where p = x/n, and x is the frequency of the observer’s retest scores that are higher (or lower) than the corresponding test scores. Significance is based on exact confidence limits for proportions when n ≤ 75 (see table B.11 in Daly and Bourke⁷).

^****Average bias: Observer ΣΔ_i/n; where Δ_i is the difference between the observer’s mean and the lead anthropometrist’s (or overall) mean measurement for the i^th participant.

^*****F ratio for bias: $\text{Observer}\text{'}\mathrm{s}\mathrm{\sum }({\mathrm{\Delta }}_i^2)/\text{Lead}\text{anthropometrist}\text{'}\mathrm{s}\mathrm{\sum }({\mathrm{\Delta }}_i^2)$ or overall means $\mathrm{\sum }(d_i^2)/(\text{same}\text{denominator}\text{as}\text{the}\text{precision}F\text{ratio})$.

^******Bias sign test: binomial proportion p, where p = x/n, and x is the frequency of the observer’s means that are above (or below) the lead anthropometrist’s mean or overall mean. Significance is based on exact confidence limits for proportions when n ≤ 75 (see table B.11 in Daly and Bourke⁷).