GROUP ASSIGNMENT

MEMBERS:
1. JERVIN MUMO KATHUMO X75/5600/2021
2. BENJAMIN MWEI X75/5561/2021
3. BILLY OMONDI OCHIENG X75/5677/2021

ANSWERS:
QUESTION 1:
Using the downloaded dataset, estimate the effect of age, sex, and marital status on
uptake of COVID-19 vaccination (20 marks)

Descriptive Statistics
Error: Reference source not found below shows the summary statistics for the key
variable used in the analysis. The average age of the household members is 24 years, with
approximately 51% female. The mean marital status is 0.525, implying that most of the
respondents were never married (coded as 0), with a max being 3 (coded as
separated/divorced/widowed). On average, each household had about 1 member being
vaccinated against COVID-19, with a maximum of 12.

Descriptive Statistics
Variable Obs Mean Std. Dev. Min Max
vaccinated dummy 1560 .592 .491 0 1
20
Age 1560 23.902 19.623 0 95
20
female 1560 .51 .5 0 1
20
Marital status 1560 .525 .826 0 3
20

The probit Regression

The probit regression analysis of vaccination status reveals several significant
demographic associations. Age shows a positive relationship with vaccination probability
(coefficient =0.01, p<0.01), indicating older individuals are more likely to be vaccinated.
Gender also plays a role, with females showing higher vaccination rates than males
(coefficient =0.054, p<0.01). Interestingly, marital status demonstrates notable negative
associations, with both married (coefficient=-0.216, p<0.01) and widowed individuals
(coefficient=-0.447, p<0.01) having lower vaccination probabilities compared to the
reference category (presumably single). While these findings are statistically significant
across a large sample of 156,020 observations, the model's explanatory power remains

1
modest with a pseudo R-squared of only 0.007, suggesting these demographic factors
explain just a small portion of the overall variation in vaccination status. Nevertheless,
the highly significant chi-square statistic (1443.405, p<0.001) confirms that the model as
a whole is superior to one without predictors.

Probit regression
Coef. St.Err. t-value p-value [95% Conf Interval] Sig
vaccinated_dumm
y
Age .01 0 35.73 0 .009 .01 ***
current marital 0 . . . . .
st~e
married -.216 .011 -20.17 0 -.237 -.195 ***
widowed -.447 .017 -26.26 0 -.48 -.414 ***
female .054 .007 8.31 0 .041 .067 ***
Constant .073 .006 11.76 0 .061 .085 ***

Mean dependent var 0.592 SD dependent var 0.491

Pseudo r-squared 0.007 Number of obs 156020
Chi-square 1443.405 Prob > chi2 0.000
Akaike crit. (AIC) 209505.655 Bayesian crit. (BIC) 209555.443
*** p<.01, ** p<.05, * p<.1

Probit regression with Robust squares accounting for multicollinearity and

heteroskedasticity
Coef. St.Err. t-value p-value [95% Conf Interval] Sig
vaccinated_dumm
y
Age .01 0 34.50 0 .009 .01 ***
current marital 0 . . . . .
st~e
married -.216 .011 -19.89 0 -.238 -.195 ***
widowed -.447 .017 -25.95 0 -.481 -.413 ***
female .054 .007 8.32 0 .041 .067 ***
Constant .073 .006 11.65 0 .061 .085 ***

Mean dependent var 0.592 SD dependent var 0.491

Pseudo r-squared 0.007 Number of obs 156020
Chi-square 1342.274 Prob > chi2 0.000
Akaike crit. (AIC) 209505.655 Bayesian crit. (BIC) 209555.443
*** p<.01, ** p<.05, * p<.1

Question 2:
Using the downloaded dataset, estimate the effect of age, sex, and marital status,
ownership of agricultural land on the uptake of usage of mobile money (20 marks).

The Descriptive statistics

This descriptive statistics table provides additional context for the population in the
previous probit regression analysis. The sample consists of 156,020 observations with an

2
average age of 23.9 years (SD=19.6), ranging from newborns (age 0) to elderly
individuals (age 95). The gender distribution is nearly balanced with 51% female
participants. Marital status has a mean of 0.525 with values ranging from 0 to 3,
suggesting it's a categorical variable with multiple codes. About 55.5% of participants
live in agricultural areas (Agriland), and mobile phone usage is high at 89.7% of the
sample. These demographics suggest a relatively young population with good technology
access despite substantial rural representation. The wide age range and diversity in living
conditions likely influence the vaccination patterns observed in the regression analysis.

Descriptive Statistics
Variable Obs Mean Std. Dev. Min Max
Age 1560 23.902 19.623 0 95
20
female 1560 .51 .5 0 1
20
Marital status 1560 .525 .826 0 3
20
Agriland 1560 .555 .497 0 1
20
Mobile phone usage 1560 .897 .303 0 1
20

The Logistic Regression

This logistic regression analyzes factors associated with living in agricultural areas
(Agriland as the dependent variable). The results show that age has a significant positive
relationship with agricultural living (coefficient=1.021, p<0.01), meaning for each year
increase in age, the odds of living in an agricultural area increase by about 2.1%. Gender
shows no significant effect on agricultural residence (p=0.514), indicating similar
rural/urban distribution between males and females. Marital status demonstrates strong
associations, with married (coefficient=0.618, p<0.01) and widowed individuals
(coefficient=0.493, p<0.01) significantly less likely to live in agricultural areas compared
to the reference group. The coefficients represent odds ratios below 1, specifically 0.618
and 0.493, indicating much lower odds of agricultural living for these groups. The model
has slightly better explanatory power than the vaccination model with a pseudo R-
squared of 0.011, though it still explains only a small portion of the variation in
agricultural residence. The large chi-square value (2270.474) confirms the model's
overall statistical significance across the 156,020 observations.

Logistic regression
Agriland Coef. St.Err. t-value p-value [95% Conf Interval] Sig
Age 1.021 0 45.20 0 1.02 1.022 ***
female .993 .01 -0.65 .514 .973 1.014
current marital 1 . . . . .

3
st~e
married .618 .011 -27.82 0 .597 .639 ***
widowed .493 .014 -25.67 0 .467 .521 ***
Constant .942 .009 -6.06 0 .923 .96 ***

Mean dependent var 0.555 SD dependent var 0.497

Pseudo r-squared 0.011 Number of obs 156020
Chi-square 2270.474 Prob > chi2 0.000
Akaike crit. (AIC) 212166.593 Bayesian crit. (BIC) 212216.382
*** p<.01, ** p<.05, * p<.1

Logistic regression with Robust squares accounting for multicollinearity and

heteroskedasticity
Agriland Coef. St.Err. t-value p-value [95% Conf Interval] Sig
Age 1.021 0 43.96 0 1.02 1.022 ***
female .993 .01 -0.65 .514 .973 1.014
current marital 1 . . . . .
st~e
married .618 .011 -27.21 0 .597 .64 ***
widowed .493 .014 -25.35 0 .467 .521 ***
Constant .942 .009 -6.04 0 .923 .96 ***

Mean dependent var 0.555 SD dependent var 0.497

Pseudo r-squared 0.011 Number of obs 156020
Chi-square 2068.945 Prob > chi2 0.000
Akaike crit. (AIC) 212166.593 Bayesian crit. (BIC) 212216.382
*** p<.01, ** p<.05, * p<.1