 As we have seen, the dashboard of the data quality tool is easy to use. A limitation of the dashboard of the application, however, is that it can only be used to review those indicators or data elements for which the dashboard has been configured. We'll next see how we can use the analysis function of the WHO data quality tool to look for outliers or inconsistencies between any data elements which have data in the DHIS too. Let's start by looking at how we use analysis consistency. When we select analysis consistency, a window appears asking us to select the type of consistency we want to analyze. Consistency over time is equivalent to what we saw on the consistency time dashboard. Whereas consistency between indicators is the equivalent of what we saw on the consistency data dashboard. For the example I will now present, we want to analyze further dropout rates and to do that, we select consistency between indicators. Then click on expected result and it's now possible to select dropout as the type of analysis. We next need to decide which two indicators will define the dropout rate. Let's look at the Penta 1 to Penta 3 dropout rate. To do that, we have to first select Penta 1 and for that we need to find out which indicator, which data element group to select. And if we select immunization, we find that Penta 1 given before 12 months of age is a data element in the immunization data element group. Next we need to select Penta 3 given before 12 months of age. Again, we need to select the data element group immunization and the immunization data element group includes Penta 3rd doses given before 12 months of age. So already you see some of the challenge of using the analysis function in that we need to configure it. And to configure it we need to know what is the data element group or indicator group we want to select from. What is the data element or indicator we want to select and so forth. This configuration requires that the user be familiar with how the data sets are organized. So this is not a function that can easily be used by the beginner. But we've now set up to look at the dropout rate between Penta 1 given before 12 months and Penta 3 given before 12 months. Now we only need to select the period. Let's select last month, last year as the period and disaggregate by district then click on analyze. The screen shows us a chart that is identical to what we saw on the consistency data dashboard. But in addition to this histogram with one bar for each district and two districts that have a negative dropout rate, we have a table. And the table shows us the number of first doses of Penta given by each district and the number of third doses given by each district. And it shows us for these two districts with negative dropout rates that the number of third doses given in the in 2019 was greater than the number of first doses. So we get this additional information when we use the analysis function. We can select one of these districts by clicking on the row and the road turns light yellow. When we do that, a table with a single row for District A2 appears at the bottom of the screen. And a chart showing for each month over the 12 month period, the value of Penta 1 and the value of Penta 3. And notice what we see. For most months, we see that District A2 reported a higher number of third doses of Penta than first doses of Penta. For example, in May of 2019, 1063 first doses, but 1139 third doses. So District A2 has a consistent practice of reporting higher numbers of third doses of Penta vaccine than first doses of Penta vaccine. And then for November of the 12 month period, there's this quite suspicious value that may be due to an error in data entry because the number of third doses is so much greater than the number of first doses. But for these other months, there seems to be some habit of reporting higher numbers of third doses, just a bit higher than first doses of Penta vaccine. We can continue the investigation by clicking on the menu icon at the end of this single row. If we're allowed to drill down, what we now see is a histogram that disaggregates the Penta 1 to Penta 3 dropout rate for 2019 by individual health facility. We see that almost all the health facilities of District A2 have been reporting higher numbers of third doses than first doses of Penta vaccine. Most all of the facilities have negative dropout rates. We can again select one of these health facilities, one of these rows and look at the trend over the 12 month period in the values for this health facility. And we see that this one facility 197 again appears to have some kind of a very suspicious value that is probably erroneous. This reported 10 times as many third doses as first doses of Penta vaccine. So in summary, by using the analysis function and looking at analysis of dropout rates. We get additional information that allows us to investigate and determine is the negative dropout rate for a district due to a single erroneous value, a single outlier such as here. Is it due to a pattern amongst many different health facilities and for many different months of reporting higher numbers of third doses than first doses. Let's look at one more example before we then see how we can use analysis to look at outliers. With this example, we're going to look at analysis between indicators, but this time, we're going to look at two indicators that we expect to be roughly equivalent. And in this case, we're going to compare third doses of Penta vaccine to third doses of OPV vaccine. These two vaccines are typically given on the same visit. So we expect the values to be roughly the same again setting period to 2019. Disadvigation to district. Click on analyze. And we see a scatterplot which reassures us that indeed for all of the districts, the value of third doses of Penta vaccine is closely related to the value of third doses of OPV vaccine. And we could if we wanted to do some investigation, although the fact that all of these dots are between the two gray lines reassures us. Let's look next at how to use the analysis function to check for outliers and missing data. The function works just as the outlier dashboard does, except it is much more flexible in that it allows us to select any data element at all, not just the ones that are configured for the dashboard. In fact, we could select from multiple data sets and multiple data element groups. We could even select all data elements from a data element group set period to last 12 months organization unit to desegregation by district. Analyze. We're waiting for the data to process. So we get the table showing us all of the suspicious values for the selected data elements. It has the same options as with the outlier dashboard so we can filter to identify only the most suspicious values, or we can filter. At facility level, we can use it to identify facilities that have large amounts of missing data. In summary, the analysis function works in a way that is quite similar to the dashboard of the WHO data quality tool, but it is much more flexible. The price that you pay for using the analysis function is that you have to configure it and to do that you need to be familiar with how the data sets are organized and you need to find just the data element or indicator that is most appropriate for the analysis.