I did a bit research on Node.js CSV package.

Use the header

If your CSV file contains a header row as supposed in comment by GrafiCode, like in this example:

"Day","Daily Device Installs"
"2021-09-15",1
"2021-09-16",1

Then CSV Parser has a feature to use the header row with column-names. See the columns option.

Benefit:

1. log the header
2. map the column-names (for simple use in code)
3. use it to make your code clean & expressive
4. defend against changes of column-order inside the input CSV

var csvData=[];
let test = 0;

// options: use default delimiter comma and map header
let parser = parse({
  columns: header =>
    header.map( column => {
      console.log(column);  
      // could also map (e.g. similar to Snake_Case)
      return column.replace(/ /g,"_");
    })
}

function addToCounter(value) {
   if (!isNaN(value))
      console.log("WARN: not a number: ", value)
      return;
   test += value;
}

// read from file
fs.createReadStream(__dirname+'/test2.csv','utf16le').pipe(parser)
   .on('data', function(csvrow) {
      csvData.push(csvrow);
      addToCounter(csvrow.Daily_Device_Installs); // the column name as mapped with underscore
   })
   .on('end',function() {
      console.log(test)
   });

Note:

• I extracted the counter-increment to a function.
• Your csvData array now contains for each row an object (with column-names as keys) instead an array of columns.

try

if (!isNaN(csvrow[2])) test += +csvrow[2];