Power of Survey Proved

Today, election result of Delhi (Indian capital city) legislative Assembly declared. Following my earlier post on "Power of Survey", the result proved the power of survey. The table below presents the comparison on exit poll prediction and actual result.

On analyzing the result with respect to projection, it can be seen only six seats deviated for two party. This is normal. It is hard to find any survey that can precisely present the result. Nor still any body can design the survey that can precisely give the result. Always the survey results are in margin of error.

Here's if you go to earlier post, you can see the table that presents result of exit poll conducted by five media houses. In that table, four media houses has presented the exit poll prediction in range. In this range AAP party actual result touched the upper range while BJP actual result touched the lower range.

So, during the sample size calculation, we usually jot down to two terms: "Level of confidence" and "Margin of Error". These two terms are most important part of any M&E survey. It is also hardest part to explain in layman language to general mass.

Statistically, if you say 95% Level of confidence, it means a range of values that you can be 95% certain contains the population mean.Lets reflect ourselves being layman, can I make any sense on this definition?

Describing margin of error is easy compare to level of confidence.A margin of error tells you how many percentage points your results will differ from the real population value. It is simply understood, if anybody said this survey result has 10% margin of error then that means the presented survey result will vary by plus minus 10% from real population value.

But, the absolute presentation of Margin of error without mentioning the level of confidence will be wrong and misguide the audience. So, most survey results are presented in .... result with 95% confidence level and 10% margin of error.

Again, reflecting from layman perspective, it is hard to understand confidence level. Should any layman understand, surveyor are 95% confidence in their survey. Or should understand I as a user/reader of survey report be confident on 95% level in the results.

Both the understanding are wrong. It should be if same survey carried out 100 times, .... result will be same for 95 times with ...... result shall vary by plus minus 10%.

The Delhi election exit poll results are analogous with "Level of confidence" and "Margin of Error". The survey (exit poll) carried out five times by five media houses with same pattern of results and actual result had touch the predicted upper and lower range.

Thus it can be said "Power of Survey again Proved".