The fly in the ointment in Region 4 was the small parties

Apr 08, 2020 Features / Columnists, Peeping Tom

---

The statement below appeared on a Facebook page. The page is purported to be connected to either the APNU+AFC or its supporters.
The statement reads as follows “A simple comparison of the final results for 2015 and 2020 shows that those published by GECOM is far more within the normal, logical expectation based on historical, demographic and statistical trends. Those published by the PPP (specifically for Region 4) run contrary to the historical voting trend, demographic trends and statistical probability.”
It would not be surprising to learn that this statement contains the reasoning behind the APNU+AFC’s justification for its contentious claim of being the victors of the 2020 General and Regional Elections. What those making this claim fail to appreciate is that simple comparisons can form the basis of false propaganda but simple comparisons are not associated with statistical probability.
Statistical probability is a scientific method. It is not opinion. It can therefore be used to test claims, which are made about the elections.
Fortunately, a local social scientist, Dr. Thomas Singh undertook to test various hypothesis contained in that statement. He has published his findings online, having summarized them in a letter published in the Stabroek News of April 5, 2020.
Dr. Singh set out to test two principal hypotheses. Firstly, whether there is any association between past elections’ results and those of the 2020 elections for Region 4. Secondly, he sought to test whether the declarations made for Region 4 are independent of each other, that is, whether the three declarations – one by the PPPC and two by Mingo – are related to each other.
The findings of his research debunk the contention that the results of elections in 2011 and 2015 can relate to the 2020 results. His findings also indicate that Mingo’s two declarations are independent of each other.
Since this column is written for a general audience, it is necessary to offer a basic explanation of how hypothesis testing works. If, for example, you wish to show a relationship between past election results and the present election results, you first have to establish the hypothesis about it being tested. In this case, it is that there is a relationship between the two sets of results.
However, in actually conducting the tests you do not set out to prove the hypothesis. Rather in order to prove the hypothesis, you have to reject the opposite of the hypothesis which is called the null hypothesis. In this instance, if you can reject the null hypothesis, that is, if you can reject that there is no relationship, then you can accept that there is a relationship.
I have reviewed the statistical tests done by Dr. Thomas Singh. I have only a slight problem with one aspect of one of his findings. His finding in relation to Mingo’s two declarations for Region 4 was interpreted by him to mean that the two declarations are “completely unrelated to each other.”
I suggest, respectfully, that the above may be too strong an interpretation given that the p-value is 0.02874. I would suggest that at the five percent significance level, you can reject the null hypothesis and thus say that the two Mingo declarations were not independent of each other.
However, at the one percent significance level, you cannot reject the null hypothesis. Thus, instead of concluding that the declarations were “completely unrelated” to each other, I would have concluded that it cannot be ruled out that they were unrelated to each other. The precision of the language is important in presenting findings of hypothesis testing.
In relation to whether past election results in District 4 are related to the 2020 results, the findings are negative. In other words, the distribution of past election results (Dr. Singh goes back to 2001) cannot be used to predict the outcome of the 2020 elections.
Having reviewed Dr. Singh’s findings on this score, I have made an interesting observation. The reason why past elections results cannot be used to predict the 2020 results for District 4, is mainly due to the improved performance of the small parties which contested that Region.
The small parties collectively gained more than 4,000 votes in Region 4. This was the fly in the ointment which mainly accounts for the finding that the results of past elections not being able to be used to predict the outcome of the 2020 polls.
Dr. Thomas Singh must be commended for the statistical work that he has done. It brings a scientific approach – and I daresay an objective focus – to resolving differences. More importantly, however, his findings provide sufficient grounds to cast doubts about the credibility of Mingo’s declarations and to justify a full recount of the votes in District 4.

(The views expressed in this article are those of the author and do not necessarily reflect the opinions of this newspaper.)

---

---