Vega lite trend indicator - kibana

Question

I am new to vega and kibana, and I am trying hard to build an indicator which shows the average monthly change of a field called 'Rating'. So far I was able to get the overal average of Rating, but I can't figure out how to get the current and the previous values so I could make a difference between them. Ideally, if the monthly change goes up we would have a '+' or an upper triangle, and the opposite if it goes down. I can't share my data, but this is my code so far (where DATE is just a datetime field, and Rating are just scores):
{
"$schema": "https://vega.github.io/schema/vega-lite/v5.json",
"data": {
"values": [
{"DATE": "2021-01-01", "Rating": 10},
{"DATE": "2021-01-02", "Rating": 8},
{"DATE": "2021-01-03", "Rating": 5},
{"DATE": "2021-01-04", "Rating": 6},
{"DATE": "2021-01-05", "Rating": 9},
{"DATE": "2021-01-06", "Rating": 7},
{"DATE": "2021-01-07", "Rating": 10},
{"DATE": "2021-01-08", "Rating": 4},
{"DATE": "2021-02-01", "Rating": 2},
{"DATE": "2021-02-02", "Rating": 1},
{"DATE": "2021-02-03", "Rating": 3},
{"DATE": "2021_02-04", "Rating": 5},
{"DATE": "2021-02-05", "Rating": 8},
{"DATE": "2021-02-06", "Rating": 3},
{"DATE": "2021-03-01", "Rating": 2},
{"DATE": "2021-03-02", "Rating": 1},
{"DATE": "2021-03-03", "Rating": 3},
{"DATE": "2021_03-04", "Rating": 5},
{"DATE": "2021-03-05", "Rating": 8},
{"DATE": "2021-03-06", "Rating": 3}
]
},
"transform": [
{"calculate": "toDate(datum['DATE'])", "as":"date"},
{"timeUnit":"quarter", "field":"date", "as":"quarter_date"},
{"joinaggregate":[{"op":"average", "field":"Rating", "as":"mean_rating_100"}],"groupby":["quarter_date"]}],
"mark":{"type":"text","fontSize":30,"fontWeight":"bold","tooltip":true},
"encoding":{
"text":{
"field":"mean_rating_100","aggregate":"average", "format":".1f"}},
"config":{
"legend":{"disable":true},
"view":{"stroke":"transparent"}}
}
Any help will be more than welcome!
Thanks!
Updated code:
Vega Editor

Answer 1

The only requirement I understand in here is that you want to calculate the previous minus current rating value. You can use a window transform where you can get the previous data. Since you are new to Vega-lite, you can view the data in the bottom part, there is a data_viewer section which will allow you to check the fields of data that gets generated via transformations (Screenshot of data_viewer)
Below is the sample snippet of a previous minus current values and also the editor. You can try performing and get the mean rating on it.
{
"$schema": "https://vega.github.io/schema/vega-lite/v5.json",
"data": {
"values": [
{"DATE": "2021-01-01", "Rating": 10},
{"DATE": "2021-01-02", "Rating": 8},
{"DATE": "2021-01-03", "Rating": 5},
{"DATE": "2021-01-04", "Rating": 6},
{"DATE": "2021-01-05", "Rating": 9},
{"DATE": "2021-01-06", "Rating": 7},
{"DATE": "2021-01-07", "Rating": 10},
{"DATE": "2021-01-08", "Rating": 4},
{"DATE": "2021-02-01", "Rating": 2},
{"DATE": "2021-02-02", "Rating": 1},
{"DATE": "2021-02-03", "Rating": 3},
{"DATE": "2021-02-04", "Rating": 5},
{"DATE": "2021-02-05", "Rating": 8},
{"DATE": "2021-02-06", "Rating": 3},
{"DATE": "2021-03-01", "Rating": 4},
{"DATE": "2021-03-02", "Rating": 6},
{"DATE": "2021-03-03", "Rating": 3},
{"DATE": "2021-03-04", "Rating": 10},
{"DATE": "2021-03-05", "Rating": 8},
{"DATE": "2021-03-06", "Rating": 3}
]
},
"width": 400,
"height": 500,
"transform": [
{"timeUnit": "month", "as": "Month", "field": "DATE"},
{
"aggregate": [{"op": "mean", "field": "Rating", "as": "avg"}],
"groupby": ["Month"]
},
{
"window": [{"op": "lag", "field": "avg", "as": "lastAvg"}],
"frame": [1, 0]
},
{
"calculate": "isValid(datum.lastAvg) ? (datum.avg === datum.lastAvg ? \"/\" : (datum.avg > datum.lastAvg ? \"+\" : \"-\")) : \"/\"",
"as": "diff"
},
{"calculate": "datum.lastAvg-datum.avg", "as": "diff2"},
{"calculate": "datum.lastAvg>datum.avg", "as": "up"},
{"calculate": "datum.lastAvg<datum.avg", "as": "down"},
{
"calculate": "if(datum.up,'🠹',if(datum.down,'🠻','🢝'))",
"as": "symbol2"
},
{"calculate": "datum.symbol2 + ' ' + abs(datum.diff2)", "as": "test"},
{"calculate": "monthFormat(month(datum.Month))", "as": "dateInMonth"},
{"filter": {"field": "dateInMonth", "equal": "March"}}
],
"mark": {
"type": "text",
"fontSize": 30,
"fontWeight": "bold",
"tooltip": true
},
"encoding": {
"text": {"field": "test"},
"y": {
"field": "Month",
"type": "nominal",
"timeUnit": "yearmonth",
"axis": {"labels": false, "ticks": false}
},
"tooltip": [
{"field": "Month"},
{"field": "lastAvg"},
{"field": "avg"},
{"field": "diff"}
]
},
"config": {"legend": {"disable": true}, "view": {"stroke": "transparent"}}
}

Answer 2

I bet window transform would be helpful on comparison between continuous data. You may share some dummy data via Vega Editor if more help is needed=)
Based on the further trial from #MTMM, I further simplify the calculations. To get the last month record, I recommend aggregation of operation argmax instead of filter because it is a bit more dynamic (in case you don't know the filtering criteria, and also treasure every opportunity to try something new~)
2021-07-22 Vega Editor
2021-07-21 Vega Editor