Table 1.
Estimating mutant frequencies using different methods: validation using a test dataset from Tanzania39. Frequencies of the 540E mutation are estimated in 24 areas in 2 different years.
| Method | Data required | Assumptions | Mean squared error | Reference for Method | ||||
|---|---|---|---|---|---|---|---|---|
| Mixed infections | Mean population MOI | Slide prevalence | Location | Assumption about detection of parasite clones | Other | |||
| 5B | √ | √ | Imperfect MOI-dependent | 0.0031 | This paper extended from32 | |||
| 5C | √ | √ | Imperfect MOI-independent | 0.0037 | This paper | |||
| 7A | √ | √ | 100% | mean MOI varied during fitting | 0.0061 | This paper extended from31 | ||
| 5A | √ | √ | 100% | 0.0062 | This paper extended from31 | |||
| 6A | √ | √ | 100% | Assume negative binomial MOI distribution | 0.0071 | This paper extended from31 | ||
| 3A | √ | √ | 100% | 0.0071 | 31 | |||
| 4A | √ | √ | 100% | 0.0083 | This paper extended from31 | |||
| 8A | √ | 100% | 0.0140 | This paper extended from31 | ||||
We compare frequencies estimated from the full dataset with information on MOI and resistance markers for each individual (Method 2, see Methods) with frequencies estimated from partial summary data using different methods. We show mean squared error to quantify the difference between the two sets of estimates in each comparison. The methods are ordered by increasing mean squared error; lowest indicates more similar estimates. The column ‘mixed infections’ indicates that data on the proportion of mixed wild type-resistance infections was used to estimate frequencies. ‘Slide prevalence’ indicates that mean MOI was estimated using the relationship in Fig. 4. ‘Location’ indicates that longitude and latitude of the survey location were used to obtain estimates of slide-prevalence from the Malaria Atlas Project, in order to estimate mean MOI using Fig. 4 45. Where imperfect detection of parasite clones was assumed, this was included when fitting to both the full and partial datasets (letters A to C after the Method number denote assumption about detection; see Methods. For example, in the first row, method 5B is compared to method 2B, in the 2nd row method 5C is compared to method 2C etc).