Statistical terms and methods
Aboriginal and Torres Strait Islander peoples and non-Indigenous population descriptors
'Aboriginal and Torres Strait Islander peoples' is the preferred descriptor used throughout the report. 'People' is an acceptable alternative to 'peoples' depending on context, but in general, the collective term 'peoples' is used. The 'Indigenous Australians' descriptor is inclusive of all Aboriginal and Torres Strait Islander groups, and is also used where space is limited.
The 'non-Indigenous' descriptor is used where the data collection allows for the separate identification of people who are neither Aboriginal nor Torres Strait Islander. The label 'other Australians' is used to refer to the combined data for non‑Indigenous people, and those for whom Indigenous status was not stated.
A crude rate is defined as the number of events over a specified period (for example, a year) divided by the total population at risk of the event.
An age-specific rate is defined as the number of events for a specified age group over a specified period (for example, a year) divided by the total population at risk of the event in that age group. Age-specific rates in this report were calculated by dividing, for example, the number of deaths in each specified age group by the corresponding population in the same age group.
Age-standardisation controls for the effect of age, to allow comparisons of summary rates between two populations that have different age structures. Age-standardisation is used throughout this report when comparing Aboriginal and Torres Strait Islander peoples with non-Indigenous Australians for a range of variables where age is a factor e.g. health-related measures. The main disadvantages with age-standardisation are that the resulting rates are not the real or 'reported' rates for the population. Age-standardised rates are therefore only meaningful as a means of comparison.
Age-standardised rates are generally derived using all age groups. However, in some cases in the Health Performance Framework report, the age-standardised rates were calculated for a particular age range to support study of a specific population group (for instance, the age-standardised data for some mortality indicators were derived for the age range 0–74).
Rate ratios are calculated by dividing the rate for Indigenous Australians with a particular characteristic by the rate for non‑Indigenous Australians with the same characteristic.
A rate ratio of 1 indicates that the prevalence of the characteristic is the same in the Indigenous and non‑Indigenous populations. Rate ratios of greater than 1 suggest higher prevalence in the Indigenous population and rate ratios of less than 1 suggest higher prevalence in the non‑Indigenous population.
Rate difference is calculated by subtracting the rate for Indigenous Australians from the rate for non‑Indigenous Australians for the characteristic of interest.
Relative standard error
Relative standard error (RSE) is a measure of sampling error, which is obtained by expressing the standard error as a percentage of the estimate.
This image is of the formula used to calculate the relative standard error of an estimate. It expresses the standard error as a percentage of the estimate in order to quantify sampling error. The standard error is divided by the estimate and this product is then multiplied by 100.
The ABS considers that only estimates with relative standard errors of less than 25%, and percentages based on such estimates, are sufficiently reliable for most purposes. Relative standard errors between 25% and 50% should be used with caution. Estimates with relative standard errors greater than 50% are considered too unreliable for general use.
The observed value of a rate may vary due to chance even where there is no variation in the underlying value of the rate. A 95% confidence interval (CI) for an estimate is a range of values that is very likely (95 times out of 100) to contain the true unknown value. CIs have not been presented for all administrative datasets as investigative work is underway into the validity of using CIs for these datasets.
Where the 95% CIs of two estimates do not overlap it can be concluded that there is a statistically significant difference between the two estimates.
As with all statistical comparisons, care should be exercised in interpreting the results of the comparison. If two rates are statistically significantly different from each other, the difference is unlikely to have arisen by chance. Judgement should, however, be exercised in deciding whether or not the difference is of any practical significance.
The standard method of calculating CIs has been used in this report. Typically in the standard method, the observed rate is assumed to have natural variability in the numerator count (for example, deaths, hospital visits) but not in the population denominator count. Also, the rate is assumed to have been generated from a normal distribution (Bell curve). Random variation in the numerator count is assumed to be centred around the true value; that is, there is no systematic bias.
The formulas used to calculate 95% confidence intervals using the standard method are:
This image is of the formula used to calculate 95% confidence intervals for crude rates. The crude rate is divided by the square root of the number of deaths or events. The product of this is then multiplied by the crude rate plus or minus 1.96.
Where d = the number of deaths or other events
This image is of the formula used to calculate 95% confidence intervals for age standardise rates. The formula is similar to that used above for crude rates, however, the formula to calculate age standardised rates is also included with in the confidence interval formula.
Where wi = the proportion of the standard population in age group i
di = the number of deaths or other events in age group i
ni = the number of people in the population in age group i
Annual change and per cent change were only calculated for series of 4 or more data points. The 95% CIs for the standard error of the slope estimate (annual change) based on linear regression are used to determine whether the apparent increases or decreases in the data are statistically significant at the p < 0.05 level. The formula used to calculate the CIs for the standard error of the slope estimate is:
This image is of the formula used to calculate 95% confidence intervals for the standard error of the slope estimate (annual change) and is used to determine whether the apparent increases or decreases in the data are statistically significant at the p
Where x is the annual change (slope estimate)
t*(n-2) is the 97.5th quantile of the tn-2 distribution.
If the upper and lower 95% confidence intervals do not include zero, then it can be concluded that there is statistical evidence of an increasing or decreasing trend in the data over the study period.
Significant changes are denoted with a * against the annual change statistics included in relevant tables.
Testing rate differences and rate ratios
If the 95% CIs of the difference in rates do not include zero, then it can be concluded that there is statistical evidence of a difference in rates. If the 95% CIs of the rate ratio do not include 1, then it can be concluded that there is statistical evidence of a difference in the rates contributing to the rate ratio.
Tables include a * next to the rate ratio and rate difference to indicate that rates for the Indigenous and non‑Indigenous populations are statistically different from each other at the p < 0.05 level (based on 95% CIs). Where results of significance testing differed between rate ratios and rate differences, caution should be exercised in the interpretation of the tests.
The word 'significant'
Statistically significant differences, for example between jurisdictions or over time, are denoted as 'significant'. The word 'significant' is not used outside its statistical context.
Significance of trends rate ratios
Significance testing of rate ratio time-trends was not done in the 2014 HPF as the accuracy of this testing may be low. In the HPF, time-series analysis use linear regression analysis to determine whether there have been significant increases or decreases in the observed rates. As rate ratios do not increase or decrease linearly, applying linear regression to rate ratios may over-estimate the significance of any changes.
Annual change and per cent change
The annual change in rates and rate differences are calculated using linear regression, which uses the 'least squares' method to calculate a straight line that best fits the data. The simple linear regression line (Y = a + bX, or 'slope' estimate) was used to determine the annual change in the data over the period.
Per cent change is calculated taking the difference between the first and last points on the regression line, dividing by the first point on the line and multiplying by 100.