My Blog List

Friday, June 20, 2014

The Charge of the Light Blue Brigade

The Twelfth Night

Dr. Lester Hazlewood-Simon

I found her at 14 past 3 in the wee hours of the morning, after the massacre of the twelfth night, naked as before she was born, awash in tears. Tears flouting gravity, ascending beyond her forehead and, on exhaustion, descending to her toes onto the floor, draining and seeping under the door. Written on the wall behind her in bright, dripping red was the reason for her torrential tears: How can we have clean, general elections without water?

I took her up, put her down on the bed and covered her nakedness with the red sheet. That was all I could find. There was a single, small blue pillow but her head needed to be flat, and her feet up, to send more red blood to her brain. Without oxygen her parched lips will turn blue, a signal, lost colour.

When she recovers I will have to explain a lot to her. One pollster had told her she will be red and another pollster held that she will be blue, after the general elections. I will tell her that polls can be wrong for many reasons. The wording and even the order of the questions can lead to error. The size of the sample and the way the sample is chosen may be faulty. Even the tone of voice of the interviewer can affect the response. She will want more details.

It is unlikely that persons will lie to one poll and tell the truth to another. The sample size may have been a critical factor to explain the consistent variance between the two polls. Let us examine these factors of sample size and response by first regarding situations that are actually opposite to how they initially appear.

In medical school we had a brilliant anatomy teacher who knew every single part of the body in fine detail. We were initially surprised to learn that he was a poor surgeon. It became obvious that his very fine, detailed knowledge prevented him from operating as quickly as the average surgeon. In effect, as perfect as he was in anatomy, very few patients attended his surgical clinic despite all the fine accolades he received from the very same patients who went elsewhere.

One of our brilliant, young, local pianists commented that he learnt so much from his piano teacher, he had immense difficulty deciding what not to play. Far too many choices came to mind during improvisation. Sometimes he had to pretend and play very simply and leave all the complicated music behind.

We all know some persons, including some politicians, who are very respectable. Yet they are so incapable of getting the job done quickly that we say yes to them and behind their backs we quietly ask or beg someone else to do the job.

When we regard the two polls, the CARUSO poll quotes a margin of error of 3% while the CADRES poll quotes a margin of error of 5%. The simple mathematics here is this. The ideal poll will access every single voter in the population. The population can be the entire constituency or the entire island. Such an ideal poll is really a census in which the margin of error will be zero per cent because you have polled every single voter. Since this is difficult or impossible for pollsters to do, they poll samples of constituencies and samples of the national population. In statistical terms, the margin of error acts as a bell curve, which means there is a point at which a large sample size becomes counterproductive to polling.

The larger the sample size, the smaller the margin of error. The smaller the sample size, the larger the margin of error. The tendency may be to get as many persons polled as possible to get a small margin of error. But hold on. What happens if there is an intrinsic or extrinsic bias in the polling population?

An intrinsic bias can mean that lots of persons like the government but are reluctant to say they will vote against the government. They are conflicted between their liking the government and thinking the opposition will be able to do things better. So they tell a little lie. They lie to both polls. However, the poll with the larger sample size will include more of these little liars, and get it wrong.

An extrinsic bias will have the same effect in that the bias is fuelled by money or gifts offered to the voters. Also, intrinsic and extrinsic biases can exist in a single voter. Look at Barbuda and the result from the CADRES poll. Landslide?

One theoretical alternative or addition to the above is if one poll had interviewers who misrepresented what persons said to them, either because the interviewers were incompetent or they were corrupted by extrinsic or intrinsic bias, or both.

The simple answer I have to give this “blue vex”, expectant, naked, dripping-wet woman, now covered in red, and “in labour”, is that bigger or more is not always better. But if I were to tell her this, she will start crying all over again. She will cry more torrential tears for five, long years with the reason for more tears emblazoned on the wall next to her bed, in dribbling ocean blue: How can you have a clean government without water?

No comments: