Table 3.
Model 1 (multinomial logit) | Model 2 (random parameter logit) | |||||
---|---|---|---|---|---|---|
Attribute (base case)a | β | SE | β | SE | SD | SE |
Main effects | Random parameters | |||||
Distribution method (Only directly to individuals) | ||||||
Deliver tests for whole household | 0.008 | 0.051 | 0.055 | 0.115 | 0.632*** | 0.054 |
Kit price (per $1 increase) | −1.273*** | 0.272 | −2.017*** | 0.400 | 1.577*** | 0.214 |
Pretest support (Information leaflet) | ||||||
Telephone helpline | 0.290*** | 0.108 | 0.415*** | 0.152 | 0.048 | 0.158 |
Face‐to‐face from distributor | −0.131 | 0.088 | −0.201* | 0.120 | 0.069 | 0.202 |
Time of operation (Monday to Friday 8 am to 4 pm) | ||||||
Monday to Friday 8 am to 4 pm + evenings and weekends | −0.008 | 0.040 | −0.032 | 0.059 | 0.036 | 0.130 |
Distributor age (below 30 years old) | ||||||
Above 30 years old | 0.008 | 0.020 | −0.016 | 0.036 | 0.258*** | 0.063 |
Distributor residence (from the same village) | ||||||
From another village | −0.116*** | 0.031 | −0.145*** | 0.052 | 0.462*** | 0.061 |
Location kit collection (collection from local clinic) | ||||||
Distributed door‐to‐door | 0.698*** | 0.219 | 1.029*** | 0.335 | 0.007 | 0.179 |
Collection from mobile testing outreach sites | −0.648*** | 0.199 | −0.970*** | 0.309 | 0.404*** | 0.100 |
Interaction effects | Non‐random parameters | |||||
---|---|---|---|---|---|---|
Household distribution×Male | −0.057*** | 0.021 | −0.078*** | 0.047 | ||
Household distribution×Age | 0.003** | 0.001 | 0.004** | 0.003 | ||
Household distribution×Non‐tester | 0.066* | 0.037 | 0.102* | 0.082 | ||
Household distribution×Self‐tester | −0.080*** | 0.028 | −0.130** | 0.064 |
Model fit statistics | ||||||
---|---|---|---|---|---|---|
Number of participants | 296 | 296 | ||||
Number of observations | 2641 | 2641 | ||||
AIC | 3488.3 | 3260.9 | ||||
AIC/N | 1.321 | 1.235 |
AIC, Akaike information criterion; HIVST, HIV self‐testing; SD, standard deviation; SE, standard error;.
aSince effects coding was applied, within each attribute, utility coefficients add up to zero, that is for two‐level attributes, the coefficient of the omitted level is the same magnitude with opposite sign. *10%, **5%, ***1% level of significance with p value.