@@ -71,8 +71,8 @@ We receive data on the following two categories of counts:
71
71
72
72
### COVID-Like Illness
73
73
74
- For a fixed location $i$ and time $t$ , let $Y_ {it}$
75
- denote the Covid counts and let $N_ {it}$ be the
74
+ For a fixed location $$ i $$ and time $$ t $$ , let $$ Y_{it} $ $
75
+ denote the Covid counts and let $$ N_{it} $ $ be the
76
76
total count of visits (the * Denominator* ). Our estimate of the COVID-19
77
77
percentage is given by
78
78
101
101
\end{aligned}
102
102
$$
103
103
104
- where $Y_ {it}$ is the observed doctor visits percentage of COVID-19 at time $t $,
105
- $\text{wd}(t) \in \{ 0, \dots, 6\} $ is the day-of-week of time $t $,
106
- $\alpha_ {\text{wd}(t)}$ is the corresponding weekday correction, and
107
- $\phi_t$ is the corrected doctor visits percentage of COVID-19 at time $t $.
104
+ where $$ Y_{it} $$ is the observed doctor visits percentage of COVID-19 at time $$ t $ $ ,
105
+ $$ \text{wd}(t) \in \{0, \dots, 6\} $$ is the day-of-week of time $$ t $ $ ,
106
+ $$ \alpha_{\text{wd}(t)} $ $ is the corresponding weekday correction, and
107
+ $$ \phi_t $$ is the corrected doctor visits percentage of COVID-19 at time $$ t $ $ .
108
108
109
109
For simplicity, we assume that the weekday parameters do not change over time or
110
- location. To fit the $\alpha$ parameters, we minimize the following convex
110
+ location. To fit the $$ \alpha $ $ parameters, we minimize the following convex
111
111
objective function:
112
112
113
113
$$
114
114
f(\alpha, \phi | \mu) = -\log \ell (\alpha,\phi|\mu) + \lambda ||\Delta^3 \phi||_1
115
115
$$
116
116
117
- where $\ell$ is the Poisson likelihood and $\Delta^3 \phi$ is the third
118
- differences of $\phi$. For identifiability, we constrain the sum of $\alpha$
117
+ where $$ \ell $$ is the Poisson likelihood and $$ \Delta^3 \phi $ $ is the third
118
+ differences of $$ \phi $$ . For identifiability, we constrain the sum of $$ \alpha $ $
119
119
to be zero by setting Sunday's fixed effect to be the negative sum of the other
120
- weekdays. The penalty term encourages the $\phi$ curve to be smooth and
121
- produces meaningful $\alpha$ values.
120
+ weekdays. The penalty term encourages the $$ \phi $ $ curve to be smooth and
121
+ produces meaningful $$ \alpha $ $ values.
122
122
123
- Once we have estimated values for $\alpha$ for the Covid counts, we obtain the
123
+ Once we have estimated values for $$ \alpha $ $ for the Covid counts, we obtain the
124
124
adjusted count
125
125
126
126
$$ \dot{Y}_{it} = Y_{it} / \alpha_{wd(t)}. $$
@@ -131,16 +131,16 @@ above.
131
131
### Backwards Padding
132
132
133
133
To help with the reporting delay, we perform the following simple
134
- correction on each location. At each time $t $, we consider the total visit
134
+ correction on each location. At each time $$ t $ $ , we consider the total visit
135
135
count. If the value is less than a minimum sample threshold, we go back to the
136
- previous time $t-1$, and add this visit count to the previous total, again
136
+ previous time $$ t-1 $ $ , and add this visit count to the previous total, again
137
137
checking to see if the threshold has been met. If not, we continue to move
138
138
backwards in time until we meet the threshold, and take the estimate at time
139
- $t $ to be the average over the smallest window that meets the threshold. We
139
+ $$ t $ $ to be the average over the smallest window that meets the threshold. We
140
140
enforce a hard stop to consider only the past 7 days, if we have not yet met the
141
141
threshold during that time bin, no estimate will be produced. If, for instance,
142
- at time $t $, the minimum sample threshold is already met, then the estimate
143
- only contains data from time $t $. This is a dynamic-length moving average,
142
+ at time $$ t $ $ , the minimum sample threshold is already met, then the estimate
143
+ only contains data from time $$ t $ $ . This is a dynamic-length moving average,
144
144
working backwards through time. The threshold is set at 100 observations.
145
145
146
146
### Smoothing
0 commit comments