Edmund Njeru Njagi,a Belgium based professor in Statistics has predicted Uhuru will win the 2017 elections by a landslide.
Using statistical techniques, Njagi found out that unless magic happens,there is no way Raila will beat Uhuru.Here is Njagi’s predictions
KENYA ELECTIONS 2017:
STATISTICAL INSIGHTS FROM 2013
PART 1
STATISTICALLY SIGNIFICANT DIFFERENCE IN CORD-JUBILEE TURNOUT
Statistical analysis of the 2013 voter turnout patterns has shown that there was a statistically significant difference in voter turnout between CORD and Jubilee inclined counties.
The odds of turning out to vote were 58% less in CORD inclined counties, in comparison to Jubilee inclined counties. This difference in odds was highly statistically significant, with a p-value of 0.0002, showing that the observed differences in turnout between these two blocks of counties were systematic (real), and not by chance.
Contrary to expectation, the level of concentration of polling stations in the counties did not show significant association with the odds of turning out to vote, after accounting for the county’s political inclination. This was demonstrated by an insignificant p-value of 0.247.
This analysis was conducted using the beta-binomial model, with the number of votes cast per county as a proportion of the county’s registered voters being the response variable, and the county’s political inclination in 2013, as well as the concentration of polling stations in the county, as explanatory variables. Other possible explanatory variables were accounted for using the beta-distributed random effect in the model.
Data on numbers of votes cast, registered voters, and polling stations, were obtained from the IEBC website. A county-specific “polling station concentration index’’ was calculated using the number of polling stations and the county’s area. Data on the latter were obtained from the website of the Kenya National Bureau of Statistics.
Further details on this analysis may be obtained from the analyst, Dr. Edmund Njeru Njagi, Assistant Professor of Biostatistics. He hopes that the series of results that he will be releasing, will help Kenyans draw relevant insights, as they move towards the elections.
KENYA ELECTIONS 2017:
STATISTICAL INSIGHTS FROM 2013 (and 2016)
PART 2
FIVE POINTS SEPARATING JUBILEE AND NASA/CORD
It would seem that only 5 points separate the Jubilee and NASA/CORD coalitions, as Kenyans gear towards the August 2017 general elections, if voting patterns in 2013, and new voter registration numbers, are anything to go by.
Statistical analysis of both the 2013 IEBC elections data, and the per-county new voter registration numbers from IEBC’s voter registration exercise conducted in early 2016, indicates that Jubilee may well get a nod from 49.40% of the total number of Kenyans projected to vote. Following closely is NASA/CORD, which seems poised to scoop 43.96% from the estimated turnout.
These percentages represent, respectively, 42.33% and 37.67% of the projected total number of registered voters.
The analysis was conducted in two stages.
At the first stage, a generalized linear mixed model was used to analyse the per-county voter turnout in 2013, to develop a predictive model for turnout in 2017. The estimated parameters in this model, including empirical Bayes estimates for the county-specific normally distributed random effect, were used to develop a predictive per-county “turnout index’’.
Counties such as Homa Bay and Nyandarua led in the turnout index, with counties such as Kilifi and Mombasa trailing at the rear of the ranking.
This predicted county turnout index was applied to the new per-county registered voter numbers, to predict county turnout in 2017, at the second stage. County voting patterns with regards to the two coalitions in 2013 were then used to provide estimates of the coalitions’ possible votes in the counties, and finally in the whole country.
Further (more advanced) statistical analyses will include, among other aspects, updating the above model and results, using the voter registration numbers from IEBC’s 2017 voter registration exercise, if and when IEBC publishes these numbers by county, on their website. To the best of the analyst’s knowledge, these 2017 per-county figures are not currently available on IEBC’s website.
More details regarding this analysis may be obtained from the analyst, Dr. Edmund Njeru Njagi [BSc. Mathematics, Statistics; Master of Statistics, Biostatistics; Doctor of Science in Mathematics, Biostatistics]; Assistant Professor of Biostatistics. As always, the analyst’s motto is: THE DATA SHALL SPEAK.
Rasta Reflections
Edmund Njeru Njagi
Edmund Njeru Njagi (Ras Njeru wa Njagi) is a Kenyan-born, Belgium-trained, London-based Assistant Professor of Biostatistics. An expert in Statistics, he is trained up to PhD level in Mathematics, Statistics, and Biostatistics. He dedicates his skills to the study of cancer.
KENYA ELECTIONS 2017:
STATISTICAL INSIGHTS FROM 2013 (and 2017)
PART 3
FIRST-ROUND WIN, OR A RUN-OFF?
Statistical analysis of the recently published data on the number of registered voters by county indicates that Jubilee and NASA are respectively poised to garner 49.55% and 43.89% of the expected total number of votes cast.
This means that depending on possible changes in turnout patterns, and political inroads, we could easily experience either a first-round win by Jubilee, or a run-off.
In a previous analysis, I demonstrated that in 2013, there was a significant difference in turnout between Jubilee and CORD/NASA inclined counties. Indeed, the odds of turning out to vote were 58% less in CORD/NASA inclined counties, in comparison to Jubilee-inclined ones, and this difference was statistically significant.
It would therefore seem that if political inroads have not been made since 2013, then NASA needs to work extra hard on turnout in its strongholds, to at least force a run-off. For Jubilee, obviously, the question would be whether the momentum that propelled its supporters to turn up in 2013 still exists.
This promises to be quite an interesting contest.
Methodology:
This statistical analysis was conducted in two stages. At the first stage, a generalized linear mixed model was used to analyse the per-county voter turnout in 2013, to develop a predictive model for turnout in 2017. The estimated parameters in this model, including empirical Bayes estimates for the county-specific normally distributed random effect, were used to develop a predictive per-county “turnout index’’.
The predicted county turnout index was applied to the recently published per-county voter registration numbers, to predict county turnout in 2017, at the second stage. County voting patterns, with regards to the two coalitions in 2013, were then used to provide estimates of the coalitions’ possible votes in the counties, and finally in the whole country.
More details regarding this analysis may be obtained from the analyst, Dr. Edmund Njeru Njagi [BSc. Mathematics, Statistics; Master of Statistics, Biostatistics; Doctor of Science in Mathematics, Biostatistics]; Assistant Professor of Biostatistics. As always, the analyst’s motto is: THE DATA SHALL SPEAK.
Credit: Edmund Njeru Njagi
