+++++++++++++++++++***************************+++++++++++++++++++++++++++

It is now 366 days since the first COVID-19 case was reported in Nigeria. As at February 28, 2021 the confirmed cases are 158,594 with 1,907 (1.23%) fatalities, however, 133,841 (85.98%) have recovered.

Based on equal days forecast, by March 01, 2022, Nigeria’s aggregate confirmed COVID-19 cases are forecast to be:

Model Confirmed cases Recoveries Fatalities RMSE
Smooth Spline 1,638,915 1,409,139 20,159 332
Upper ARIMA 814,409 700,228 10,017 213
Quadratic Polynomial 789,941 679,191 9,716 356
Essembled based on weight 25,167 21,639 310 182
Essembled based on summed weight -140,682 -120,959 -1,730 188
Lower ARIMA -558,388 -480,102 -6,868 213
Essembled based on weight of fit -721,936 -620,721 -8,880 269
Essembled with equal weight -1,403,659 -1,206,866 -17,265 226
With knots -2,638,910 -2,268,935 -32,459 190
Without knots -3,290,728 -2,829,368 -40,476 202

Refer to Table 2 and Table 3 as well as Fig. 18-20 for more details on how the estimates and forecasts were obtained.

+++++++++++++++++++***************************+++++++++++++++++++++++++++

The visuals below supports this facts, take a look!

Fig. 1a Daily observed cases of COVID-19 in Nigeria

Fig. 2 Perecentages and previous days differences of COVID-19 in Nigeria Starting from February 29, 2020 to February 28, 2021

Fig. 3 Cases recorded in percentages Starting from February 29, 2020 to February 28, 2021 (legend as Fig. 2)

Fig. 4 Cumulative cases and Forecast of COVID-19 cases in Nigeria Starting from March 01, 2021 to March 01, 2022

Fig. 4a Components of COVID-19 cases in Nigeria between February 29, 2020 and February 28, 2021

Fig. 5 Number of days since average recorded cases exceeded one in each State

Fig. 6 Daily confirmed Coronavirus cases, by number of the days since 1 first daily case recorded

Fig. 7 Number of days from February 29, 2020 to February 28, 2021 that cases were recorded in the States

Fig. 8 Number of days the last COVID19 case was recorded as at February 28, 2021

Fig. 9 Diverging Bars of COVID-19 cases in the States (normalised)

Fig. 10 Monthly summary of COVID19 cases in the States

Fig. 11 Daily recorded cases of COVID19 in the States

Fig. 12 Number of days since average recorded cases exceeded one in each geo-political zone

Fig. 13 Number of days from February 29, 2020 to February 28, 2021 that cases were recorded in the Zones

Fig. 14 Monthly summary of COVID19 cases in the Zones

Table 1 Regression estimates of the effect of price of PMS and rate of inflation on COVID-19 and vice versa

COVID19PMSInflationTeledensity
(Intercept)-83.432 ****-304.435 ****32.323 *** 84.250 ****
(12.712)    (38.822)    (7.523)    (3.013)    
PMSprice-0.254 ****         0.111 ****0.244 ****
(0.041)             (0.016)    (0.030)    
Inflation2.000 ****7.568 ****         -1.691 *** 
(0.396)    (1.110)             (0.442)    
Teledensity0.979 ****3.607 ****-0.366 ***          
(0.154)    (0.446)    (0.096)             
log(MonthCase)         -3.205 ****0.369 ****0.835 ****
         (0.511)    (0.073)    (0.131)    
nobs13         13         13         13         
r.squared0.881     0.964     0.910     0.941     
adj.r.squared0.842     0.952     0.880     0.921     
sigma1.129     4.013     0.485     1.043     
statistic22.292     79.764     30.309     47.702     
p.value0.000     0.000     0.000     0.000     
df3.000     3.000     3.000     3.000     
logLik-17.638     -34.120     -6.659     -16.606     
AIC45.276     78.240     23.317     43.212     
BIC48.100     81.065     26.142     46.037     
deviance11.480     144.939     2.120     9.795     
df.residual9.000     9.000     9.000     9.000     
nobs.113.000     13.000     13.000     13.000     
**** p < 0.001; *** p < 0.01; ** p < 0.05; * p < 0.1.

Fig. 15 COVID-19, inflation, PMS price and Teledensity in Nigeria

Fig. 15a Relationship between COVID-19, inflation, PMS price and Teledensity in Nigeria

Fig. 15b Relationship between PMS price, inflation, COVID-19 and Teledensity in Nigeria

Fig. 15c Relationship between Inflation, PMS price, COVID-19 and Teledensity in Nigeria

Fig. 15d Relationship between Teledensity, Inflation, PMS price and COVID-19 and Teledensity in Nigeria

Table 2 The coefficient estimates of the various models for forecasting of COVID-19 for Nigeria1

Spline with knotsSpline without knotsARIMAQuardratic polynomial
(Intercept)-145.569 **  -3.665          185.138 *** 
(69.142)    (78.101)         (56.300)    
bs(niz[, 1], knots = NULL)11221.969 ****                       
(199.868)                           
bs(niz[, 1], knots = NULL)2-473.447 ****                       
(127.055)                           
bs(niz[, 1], knots = NULL)31564.786 ****                       
(109.099)                           
bs(niz[, 1], knots = BREAKS)1         6.300                   
         (151.036)                  
bs(niz[, 1], knots = BREAKS)2         11.918                   
         (97.894)                  
bs(niz[, 1], knots = BREAKS)3         845.836 ****              
         (112.376)                  
bs(niz[, 1], knots = BREAKS)4         167.924 *                 
         (95.772)                  
bs(niz[, 1], knots = BREAKS)5         207.238 *                 
         (106.946)                  
bs(niz[, 1], knots = BREAKS)6         -86.896                   
         (105.882)                  
bs(niz[, 1], knots = BREAKS)7         2015.112 ****              
         (123.907)                  
bs(niz[, 1], knots = BREAKS)8         1027.205 ****              
         (124.250)                  
bs(niz[, 1], knots = BREAKS)9         342.845 ***               
         (124.117)                  
ar1                  1.004          
                  (0.114)         
ar2                  -0.300          
                  (0.063)         
ma1                  -1.654          
                  (0.106)         
ma2                  0.755          
                  (0.086)         
Day                       -0.893     
                       (0.708)    
I(Day^2)                       0.009 ****
                       (0.002)    
nobs366         366         365     366         
r.squared0.458     0.800          0.379     
adj.r.squared0.453     0.795          0.375     
sigma334.083     204.697     214.324 357.065     
statistic101.793     157.967          110.618     
p.value0.000     0.000          0.000     
df3.000     9.000          2.000     
logLik-2644.289     -2461.942     -2475.578 -2669.143     
AIC5298.578     4945.885     4961.156 5346.286     
BIC5318.091     4988.814     4980.656 5361.897     
deviance40403365.217     14916722.686          46280768.017     
df.residual362.000     356.000          363.000     
nobs.1366.000     366.000     365.000 366.000     
**** p < 0.001; *** p < 0.01; ** p < 0.05; * p < 0.1.

Table 3 Various selection criteria for the estimated models2
Spline with knots Spline without knots Smooth spline ARIMA Quardratic polynomial
Absolute Error 85246 38683 35300 40592 102747
Absolute Percent Error Inf Inf Inf Inf Inf
Accuracy 0 0 0 0 0
Adjusted R Square 0.45 0.79 0 0 0.38
Akaike’s An Information Criterion AIC 5299 4946 0 4961 5346
Allen’s Prediction Sum-Of-Squares (PRESS, P-Square) 0 0 0 0 0.37
Area under the ROC curve (AUC) 0.01 0.01 0.01 0.03 0.23
Average Precision at k 0 0 0 0 0
Bias 0 0 0 2.8 0
Brier score 110392 40756 36191 45307 126450
Classification Error 1 1 1 1 1
F1 Score 0 0 0 0 0
fScore 0 0 0 0 0
GINI Coefficient 1 1 1 1 1
kappa statistic 0 0 0 0 0
Log Loss Inf Inf Inf Inf Inf
Mallow’s cp 4 10 0 0 3
Matthews Correlation Coefficient 0 0 0 0 0
Mean Log Loss Inf Inf Inf Inf Inf
Mean Absolute Error 233 106 96 111 281
Mean Absolute Percent Error Inf Inf Inf Inf Inf
Mean Average Precision at k 0 0 0 0 0
Mean Absolute Scaled Error 1.8 0.82 0.75 0.86 2.2
Median Absolute Error 174 55 42 55 235
Mean Squared Error 110392 40756 36191 45307 126450
Mean Squared Log Error NaN NaN 0.18 0.19 2.6
Model turning point error 175 174 171 243 184
Negative Predictive Value 0 0 0 0 0
Percent Bias NaN NaN NaN NaN -Inf
Positive Predictive Value 0 0 0 0 0
Precision NaN NaN NaN NaN NaN
R Square 0.46 0.8 0 0 0.38
Relative Absolute Error 0.7 0.32 0.29 0.33 0.84
Recall -99 -2.1 -0.16 0.5 180
Root Mean Squared Error 332 202 190 213 356
Root Mean Squared Log Error NaN NaN 0.43 0.43 1.6
Root Relative Squared Error 0.74 0.45 0.42 0.47 0.79
Relative Squared Error 0.54 0.2 0.18 0.22 0.62
Schwarz’s Bayesian criterion BIC 5318 4989 0 4981 5362
Sensitivity 0 0 0 0 0
specificity 0 0 0 0 0
Squared Error 40403365 14916723 13245903 16582438 46280768
Squared Log Error NaN NaN 66 69 939
Symmetric Mean Absolute Percentage Error 0.7 0.39 0.35 0.4 0.82
Sum of Squared Errors 40403365 14916723 13245903 16582438 46280768
True negative rate 0 0 0 0 0
True positive rate 0 0 0 0 0

Fig. 18 Models of COVID-19 Cases using ensemble technology

Fig. 19 Equal duration forecast of COVID-19 Cases from the ensemble models using a native plotting method

Fig. 20 Forecast of COVID-19 cases in Nigeria Starting from March 01, 2021 to March 01, 2022 using a more advaced plotting method

Fig. 20a Lagged forecasts (1-14 days) of COVID-19 cases in Nigeria

Table 4 Summary of the dynamic COVID-19 forecasts
Model Feb 15, 21 - Feb 01, 22 Feb 22, 21 - Feb 15, 22 Feb 25, 21 - Feb 21, 22 Feb 26, 21 - Feb 23, 22 Feb 27, 21 - Feb 25, 22 Feb 28, 21 - Feb 27, 22
Without knots 2159174 1970827 1841670 836016 1763019 1704146
Smooth Spline -4624513 -3208868 -3150812 -2926571 -2740521 -2985142
With knots -2102613 -1955283 -2090525 -1908254 -1774546 -2172853
Quadratic Polynomial 861937 855922 833476 827307 820277 806383
Lower ARIMA -323057 -373517 -453547 -463721 -471315 -513812
Upper ARIMA 897077 884800 857295 854475 852839 832135
Essembled with equal weight 274067 264413 -1080801 -904653 -815313 -966913
Essembled based on weight 183750 190941 192792 192330 192393 194047
Essembled based on summed weight 283614 294141 295864 297176 299728 296197
Essembled based on weight of fit -326502 -313234 -500539 -389962 -336686 -436875
Table 3 RMSE of the models in the dynamic forecasts
Model Feb 15, 21 - Feb 01, 22 Feb 22, 21 - Feb 15, 22 Feb 25, 21 - Feb 21, 22 Feb 26, 21 - Feb 23, 22 Feb 27, 21 - Feb 25, 22 Feb 28, 21 - Feb 27, 22
Without knots 281.83 303.91 315.53 318.55 321.64 326.67
Smooth Spline 201.25 203.16 202.56 202.46 202.3 202.07
With knots 190.16 191.4 190.93 190.78 190.39 190.26
Quadratic Polynomial 346.42 348.08 350.31 350.79 351.39 353.23
Lower ARIMA 212.82 213.53 213.19 212.92 212.64 212.77
Upper ARIMA 212.82 213.53 213.19 212.92 212.64 212.77
Essembled with equal weight 216.85 221.46 223.28 223.69 224.04 225.07
Essembled based on weight 472.76 474.54 473.04 472.6 472.09 471.43
Essembled based on summed weight 470.79 472.78 471.43 471.04 470.6 470.03
Essembled based on weight of fit 288.34 285.37 280.55 279.17 277.64 275.01

Fig. 21 Percentage of COVID19 cases that resulted into casualty per State as at February 28, 2021

Fig. 22 Percentage of COVID19 cases that recovered per State as at February 28, 2021

Fig. 23 Percentage of recoveries and deaths from COVID19 cases per Zone as at February 28, 2021

Fig. 24 Distribution of COVID19 in the States as at February 28, 2021

  1. Note that Smooth spline does not estimate coefficients and other model-characteristic statistics↩︎

  2. Multiclass Area Under the Curve (MAUC) is the mean of the various AUC estimated from the same data↩︎