search

ICPI / DIV

Organizational Material
10 stars 2 forks source link

TIPS: PEPFAR Analytic Workspaces (PAWS) #127

Open jb3436 opened 5 years ago

jb3436 commented 5 years ago

Share Lessons Learned

Purpose: To document lessons learned around the 4 areas below for analysts to have a place to reference helpful tips as folks develop dossiers.

Resources Shared by DataStrong

Shared via email: Panorama development second system check in 7/8 Finer Age Bands Overview Visual Crossing Intro Analytic Workspaces FAQ - How to login to Analytic Workspaces

Project Management

Dataset and Metrics

Software Specific

Site Level Analytics

Analytic Workspaces FAQ - How to login to Analytic Workspaces.pdf

jb3436 commented 5 years ago

Copying Suggestions over from #all-paws-ondeck slack channel here before I archive it

All Panorama Dossiers are in PAWS in Reference Panorama Folder

jb3436 commented 5 years ago

Quick question: As you know the dossier is currently in two separate dossiers, one pointing to the dynamic datasets and the other to another back-end dataset. Can the two dossiers be merged into one, still retaining their datasets? Thanks a ton!

Questions from Related to HRH Dossier #3 A few requests that are still outstanding for the dossier [human resources for health 3], as discussed in our meeting yesterday, are:

Why are trend lines apparently not showing up on any bubble or scatter graphs? And why do some types of trend lines attempt to compress the Y-axis (but still not appear)

Is there a way to enable the creation of a custom attribute, based on whether an observation (such as a site) falls “above” vs. “below” a specified value for a calculated metric? I realize that this works for “regular”/“official” metrics and we could keep adding to the Metrics dataset, but the former would be a far more extensible long-term solution for us and for your company.

The ideal situation we discussed was adding a slider filter for a cutoff value. Example: allow the user to set the cutoff to 90% proxy retention, then the x-axis would automatically be split between sites that fall below 90% for retention would appear in the first chart category and sites above 90% would appear in the second chart category, all for the same graph. Right now a slider filter appears to be two-sided

The overall performance of the dossier remains puzzling. When I first created with 5 chapters it ran perfectly fine as a global tool (site level data for all OUs). Now with only 1-2 OUs it takes forever to work or manipulate anything.

Pro tip: I strongly suggest that any dossiers with site -level data have an OU prompt on the front end to reduce the data. A dossier that is site-level and includes all OU is bound to have performance issues and will also consume a great deal of system power.

You can create a new filtered attribute that has quantitative filters, or you can use a case statement I believe if you need to define multiple values for the attribute. To have a dynamic cutoff value, I think you could make a workaround where if you have an attribute that says 90, 80, 70 etc and then you make that attribute a selector. Then you create a reference line based on that selector indicator. Or you create a new attribute with a case statement based on that 90, 80, 70 attribute. In other words you can choose whether X is 90, 80 or 70. And then in your case statement you are saying: if retention < X then Group 1.

The 90, 80, 70 attribute would have to be created by Janice in the levels dataset.

jb3436 commented 5 years ago

Label in Visuals

Add a label of which OU is selected into the visualization This only seems to be availabel in "vitara charts" as part of the small multiples How can we include this on other visuals image

jb3436 commented 5 years ago

HTML Code used in Pano Dossiers

Dossier Name

Paragraph...

This dossier includes the following chapters with tables that will allow you to view your data:

  • Name of chapter
    • Interactive map that allows you to view OU data down to facility level
    • Semi annual results will be shown ....
    • The numbers displayed show the cascade going from...
    • In future reporting cycles we will be able to align ...r
  • Name of chapter
    • There are three types of treatment for precancerous lesions ....
    • explanation...

Filtering your data

  • A filter in a chapter filters the data in all the visualizations within the chapter.
  • The data in each chapter is filtered independently of the data in other chapters in the dossier.
jb3436 commented 5 years ago

Considerations for Building in DEV directly

Building in Dev directly didn't work initially due to user permissions issues. User permissions were updated and additional users will be given access to dev so they can see the dossier development

Issue: If you are having trouble adding metrics in DEV, Solution: select to keep the changes local to the dossier when editing the dataset, then you will be able to add all the metrics needed. This suggestion is only for DEV and this is not a solution for PAWS prod.

jb3436 commented 5 years ago

Multiple People working on a Dossier

Question: have users faced issues when multiple users work in the same dossier and try to save their work simultaneously? is there a way to avoid over-writing each other's work? Answer: More than one person making changes to the same object at the same time is definitely a problem (and a potential cause of metadata corruption).

Question: If not can we work in different dossiers and then merge them later? Answer: Yes

jb3436 commented 5 years ago

Color Palettes

For Microstrategy out of the box visuals (OOTB) ues below

image

For Vitara Charts Use below

image

jb3436 commented 5 years ago

Fine Age 2019 vs. TrendsFine Age 2019 Question

Example:

jb3436 commented 5 years ago

Swapping Out Derived Metrics for Systems Metrics

Once systems metrics are created, you must swap your derived metrics for the systems metrics. This should not take too long and it is important to build first and make sure the visuals function instead of waiting for the metrics.

jb3436 commented 5 years ago

Geo spatial Relationship Thoughts

From Jacob 7/11 RE: Pano Requirements and development milestones

Important point for all dossiers: I think we need to be clear on “geospatial relationship” which is probably better said as “geospatial dynamic” or some such (Nate can please chime in here) Everything can and will be map-able in the mapping dossier. There needs to be a particular justification to add a site-level map to a dossier, since that will add duplication, complexity and (could impact performance in select cases), and this justification should include a geospatial dynamic inherent in the data at the site level. I would ask Nate to weigh in, but I would say:

Geospatial dynamic does NOT mean: • It can be mapped; that is, data are associated with a geospatially definable position. • Asymmetries can be identified in the geospatial distribution of data at the site-level of granulrity. That’s all data. No data are perfectly evenly distributed geographically. You could create a map of sites in which the color of site markers is determined by whether or not their reported TX_CURR value contains the number “4” (red) or contains the number “3” (blue). You will find, by chance, some places on the map that is asymmetrically red or blue. However, there is no relationship between these data and anything that is geospatially-specific at a certain (site) level of geospatial granularity. There is nothing about the sites’ geospatial position (that is to say, its geospatial relationship/proximity to natural environmental, built environmental, animal, demographic, cultural, governmental, economic, etc. etc. characteristics/facts) that influences the presence of a 4 or a 3 in their TX_CURR value.

As best as I can define it as a layperson, geospatial dynamic means that: • The data can be influenced by the associated geospatial position (its geospatial relationship/proximity to natural environmental, built environmental, animal, demographic, cultural, economic, etc. etc. characteristics/facts) a site-level of geospatial granularity.
• For example, we map recent infections at the site level because HIV transmission is geospatially specific in so far as it involves human interaction which is spatially bound and occurs in a spatial plane (I can’t get it via SMS), and because it is related to geospatially-specific characteristics/facts (culture, economic activity, etc.) at a certain (site) level of granularity. • For a second example: in my negative example above I was going to say that there is no relationship between geospatial position and the number of vowels in a site’s name, but I realized that is false. Culture and language are geospatially specific, and they could be associated with the number of vowels in a site’s name.

This is all to say, I don’t see the geospatial dynamic in HRH distribution. HRH distribution could be related to IM, and IMs work in certain sites. Or it could be related to SNU, and SNUs contain sites. In those cases, we can have a graph showing HRH by IM or SNU1. But I don’t see how HRH allocation has a geospatial relationship/proximity to natural environmental, built environmental, animal, demographic, cultural, governmental, economic, etc. etc. characteristics/facts at a certain (site) level of granularity.

jb3436 commented 5 years ago

Filtering visuals to exclude outliers

Travis to do a demo

image

image

jb3436 commented 4 years ago

Sorting values

first create a grid, THEN convert it to your visual

Info From Microstrategy: https://doc-archives.microstrategy.com/producthelp/10.9/WebUser/WebHelp/Lang_1033/Content/advanced_sort_editor.htm

hessionra commented 4 years ago

As you come up with solutions - paste them HERE and we will update them in GitHub https://github.com/ICPI/DIV/issues/127 View in Slack

jb3436 commented 4 years ago

Converting an attribute to a metric

image

image