Monday, February 12, 2007

An Example of Sampling Bias in Dyslexia Reporting

Pre selecting has a long history in dyslexia research. " dyslexics are of average or above average intelligence." is a self fulfilling prophecy because if you look for dyslexics from the group of people having average to above average intelligence to the exclusion of below average intelligence then that is what you find. This technique produces what is called a sampling bias.

The following is from an article that suggests that schools need to do a better job teaching dyslexics how to read. That is certainly true. I just do not believe that the example below supports what is inferred from the rest of the article.

"Robert Broudo, the Landmark School's headmaster, said the school selects students with strong intellectual ability, even if they have serious reading deficits.
It then gives them concentrated help with their reading difficulties, he said, including one-on-one tutoring each day and grouping students by reading ability in each of their classes.
The results? About 90 percent of Landmark's graduates go on to college, and all of them passed Massachusetts' state graduation exam last year, compared with 65 percent of special education students elsewhere in the state. "

Let me take the above information and look at it differently to see why I suggest the results might not be what it suggests. I will use a group size of 100 as an example.

1) Normally 65 special ed students pass the Massachusetts' state graduation exam out of a group of 100.

2) Normally 35 special education students do not pass the Massachusetts' state graduation exam college out of a group of 100.

3) If you look at the top 50 of the group as indicated by intellectual ability you would expect a higher % to pass the Massachusetts' state graduation exam .

4) If you look at the bottom 50 of the group as indicated by intellectual ability you would expect a smaller % to pass the Massachusetts' state graduation exam.

5) All 50 passed the Massachusetts' state graduation exam from the top group after intervention.

6) If 15 out of 50 passed the Massachusetts' state graduation exam from the bottom group ,a 30% average, then the total would be the same.

Is it reasonable to suspect that the rate of passing the Massachusetts' state graduation exam is 100% from the top half and 30% from the bottom half for special education students when the average for the whole group 65% without the intervention? I believe it is.

For the intervention to not be indicated as any success by the standard of passing the Massachusetts' state graduation exam , the only thing that would have to be true is that the 35 special education students who did not pass the Massachusetts' state graduation exam all were from the bottom half of the students.

I think it is reasonable to suspect that the 35 students who were not going to pass the Massachusetts' state graduation exam were indeed from the bottom half of the group along with 15 students who were going to pass the Massachusetts' state graduation exam.

Do I think the intervention was of no value? NO ! Was any information given that would support the value of the intervention? NO !

How reading is taught needs to be improved, examples are everywhere in the statistics.That is why I will not say that the intervention had no value. This sloppy example does nothing to give any indication of what type of intervention works or does not work. The standard used, that the selected students did better than average as indicated by passing the Massachusetts' state graduation exam, could have also resulted from the sampling bias of selecting the better students and then comparing the selected group to the average.

1 comment:

Laqueta said...

Keep up the good work.