The decision tree might not be yet fully complete - review all possible branches of sampling assumptions setting

[Edouard-Legoupil]

The core tools behind the app is the sampling tree - which is under the module https://github.com/unhcr-americas/rmsSampling/blob/master/R/mod_screener.R -

The sampling tree is indeed a directed graph.

The initial documentation was turned into a DiagrammeR::grViz object defined as below:

#install.packages("DiagrammeR")

DiagrammeR::grViz(
  "digraph flowchart {
  node [fontname = Lato];
  edge [fontname = Lato];

 # Decision Points  #########
 facecondition [  label = 'Q: There is possibility to do \n face to face interview';
    shape = diamond;    color = '#FAEB00';   style = rounded; ];

 lessthan5000 [ label = 'Q: The population group is \n smaller than 5,000 individuals';
    shape = diamond; color = '#FAEB00'; ];

  availablereg [ label = 'Q: Up to date and \n complete registration list';
    shape = diamond;  color = '#FAEB00';  ];

  budget [ label = 'Q: Specific budget and time available \nto conduct a listing exercise ';
    shape = diamond; color = '#FAEB00'; ];

 traceable [ label = 'Q: Population group traceable';
  shape = diamond;  color = '#FAEB00'; ];

  spread [ label = 'Q: Geographic distribution \n of that population group in the country ';
    shape = diamond; color = '#FAEB00';  ];

  cluster [ label = 'Q: Do we have already  \n population size breakdown by area within the country? ';
    shape = diamond;   color = '#FAEB00'; ];

   strata [ label = 'Q: Operation need \n from statistically reliable disaggregation ';
    shape = diamond;   color = '#FAEB00'; ];

 gather [ label = 'Q: Population tend to gather at \n a certain location on a specific day/time ';
    shape = diamond; color = '#FAEB00'; ];

  network [ label = 'Q: Population is a well-connected community,\n where people tend to know each others ';
    shape = diamond;  color = '#FAEB00'; ];

   expert [ label = 'Q: Adequate time and expertise to \n carry out Respondent Driven Sampling';
    shape = diamond; color = '#FAEB00'; ];

  # Method Statement #########
  nonprob [ label = 'Non-probabilistic methods';
    shape = rect;    color = '#99c7e4';];

  srs [ label = 'Simple Random Sampling \n(SRS) without stratification';
    shape = rect;   color = '#338ec9';];

  ppsframe  [ label = 'Multiple-Stage \nCluster Sampling\n with Frame';
    shape = rect;  color = '#338ec9'; ];

  srsstrata  [ label = 'Simple Random Sampling\n (SRS) within Strata';
    shape = rect; color = '#338ec9'; ];

  acs [ label = 'Adaptive Cluster Sampling';
    shape = rect; color = '#338ec9'; ];

  lts [ label = 'Time Location Sampling';
    shape = rect; color = '#66aad7'; ];

  rds [  label = 'Respondent Driven Sampling\n (RDS)';
    shape = rect; color = '#66aad7'; ];

  # Edge statements  #########
  lessthan5000 -> facecondition;
  lessthan5000 -> nonprob[ label = 'Yes' ];
  facecondition ->  availablereg;
  availablereg -> traceable[ label = 'Yes' ];
  traceable -> budget[ label = 'No' ];
  traceable -> spread[ label = 'Yes' ];
  budget -> nonprob[ label = 'No' ];
  budget -> strata[ label = 'Yes' ];
  spread -> srs[ label = 'Concentrated' ];
  spread -> budget[ label = 'Scattered' ];
  strata ->  srsstrata[ label = 'Yes' ];
  strata ->  cluster[ label = 'No' ];
  cluster ->  acs[ label = 'No' ];
  cluster ->  ppsframe[ label = 'Yes' ];
  availablereg ->  gather[ label = 'No' ];
  gather -> lts[ label = 'Yes' ];
  gather -> network[ label = 'No' ];
  network -> expert[ label = 'Yes' ];
  network -> nonprob[ label = 'No' ];
  expert -> rds[ label = 'Yes' ];
  expert -> nonprob[ label = 'No' ];

} ")

However the current decision flow might still not be smooth:

• How to integrate the data collection mode decision in the sampling decision flow?
• Is the order of decision ok?

[almouthnna]

sometimes people mix between stratum and cluster, i you need to identify metadata,