Stanines Versus Average Bands: Some Clarification

The Maine stanines were collected in 1995-1996 in all Reading Recovery (RR) schools. They reflect a stable estimate of first grade performance on the Observation Survey because the sample of over 1000 first graders on which the stanines are based was randomly selected from all first graders, including RR children.

• Performance is categorized into stanines 1 (low) through 9 (high) with the average performance falling in stanine 5.
• The stanines are useful statewide benchmarks. The achievement goal of first graders is stanines 5 and up, which is the average performance or better.
• Fall stanines can also be useful statewide benchmarks to assess performance at the end of kindergarten, while year-end stanines can also be useful statewide benchmarks to assess performance in the fall of second grade.

The Maine average bands are yearly statewide performance estimates of first graders since 1991-1992. Each average band is based on a new sample.

• The first graders are randomly selected from every first grade child at each Reading Recovery school. A child who receives Reading Recovery is excluded afterwards.
• The yearly random sample serves as a comparison group for children who receive Reading Recovery services, since the performance goal of Reading Recovery children is to meet the performance of their typical first grade peers.
• The performance of the random sample varies between districts and even between schools in the same district, so it is important to compare the progress of Reading Recovery children with the random sample comparison group at their own school.
• Recognize that the sample sizes for the average bands change each year. The number of schools implementing Reading Recovery since 1991-1992 has increased considerably, rendering a larger random sample each year.

How the Reading Recovery Average Band is Computed

The average band for Reading Recovery is a measure of typical performance of first graders. This is because it represents the average scores on the Observation Survey of a randomly selected group of children.

Each fall, Reading Recovery schools randomly select first graders to be tested. If a child selected is a first-round Reading Recovery child, the child’s name is not accepted and the random selection process continues until enough children are selected for the random sample comparison group. By including every first grader including first-round Reading Recovery children in the selection process the sample is more likely to be a representation of first grade performance.

Children who are a part of the random sample at the beginning of the year who then get picked up for Reading Recovery services in the 2nd or 3rd round are removed from the random sample. Attrition of the sample is expected and is a natural part of data collection. This is why it is always better to have more rather than less children selected in the fall for the original random sample.

To Calculate an Average Band

In your state data packet for 1999-2000, look at Tables 3.3a to f. Notice the bottom row of each table represents the scores of the random sample comparison group. Note that there are two columns of interest for this calculation: the mean and the standard deviation.

To calculate the year-end text reading average band, for instance, we would look at the year-end mean and standard deviation of the random sample comparison group. In 1999-2000 the mean was 21.8, as seen in Table 3.3a, and the standard deviation (SD) was 6.8. If you picture the distribution of scores of the random sample, imagine that the standard deviation is the middle range, half being above the mean, and half being below the mean:

|<--------- SD=6.8 --------->|

<---- 3.4 ----><---- 3.4 ---->

<-------Average Band ------->

<------------------------------------------|------------------|------------------|------------------------------------------>
Mean = 21.8

Therefore the year-end text reading average band is the mean plus or minus half of the standard deviation. So 21.8 minus 3.4 is 18.4, the lower end, and 21.8 plus 3.4 is 25.2, the upper end. The only scores that fall between 18.4 and 25.2 are text reading level 20-24, so the year-end text reading average band for Maine in 1999-2000 is text reading level 20-24.

You can apply the same process to calculate the average band of any sample if you have the mean and standard deviation. Keep in mind that bands calculated on small samples have questionable validity and therefore are unreliable.