MYP Design Collecting, Storing and Presenting Data Notes

Quantitative and Qualitative Data Answer Different Questions

1. Quantitative data is anything you can count or measure: 47 responses, 12.4 seconds, a rating of 4 out of 5, a mass of 230 g.
2. Qualitative data is anything captured in words or images: what a user said about the handle, a photo of where the strap wore through, a note that three people held the product upside down.
3. Asking users to rate the grip from 1 to 5 gives you something to average, and asking what they disliked gives you the reason that average came out at 2.
4. Primary data is what you gather yourself by testing, watching or asking, while secondary data comes from someone else's work and has to be credited to them.
5. Strong research in Inquiring and analysing usually mixes both: thirty quick responses to find the pattern, four proper conversations to explain it.

A Spreadsheet Is a Data Store With Rules

1. A spreadsheet becomes useful the moment you treat it as a proper table: one row per thing, one column per property, one header row, no blank rows in the middle.
2. Keep every column to a single data type, so a column called mass_g holds 230 and never holds "230 g" or "about 230".
3. Do not merge cells inside a data range, because merged cells break every sort, filter and formula that touches them.
4. Use short lower case column names with no spaces, such as user_id, test_date and grip_score, so formulas stay readable.
5. Keep the raw responses on one sheet and your workings on another, so you can always get back to exactly what people said.
6. Once your data splits into several linked tables, such as users, tests and products, you have outgrown a spreadsheet and moved into database territory, which the article on database structure deals with.

Sorting and Filtering: Finding the Answer Inside 300 Rows

1. Sorting puts rows in order by one column, so sorting by rating lifts your worst results to the top where you have to look at them.
   1. Always sort the whole table rather than one column on its own, or you will end up with the wrong mass beside the wrong sample.
2. Filtering hides the rows that do not match a condition while leaving the full table underneath.
3. Filter to ages 11 to 14 who scored the grip 1 or 2 and you are looking at the exact group whose problem your design has to solve.
4. A pivot table counts and groups for you, giving how many of each rating broken down by year group in about four clicks.
5. Conditional formatting colours any cell that meets a rule, which makes odd readings visible before you go hunting for them.

Four Formulas Do Most of the Work

1. =AVERAGE(B2:B31) gives the mean, which is the right summary when results cluster and the wrong one when a single reading is extreme.
2. =MEDIAN(B2:B31) gives the middle value, and it survives one mistyped 900 sitting in a column where everything else is near 20.
3. =COUNTIF(C2:C31,"yes") counts the rows matching a condition, which is how thirty answers become a percentage you can quote.
4. =MAX and =MIN give you the range, and range matters more than the average for anything about fit, reach or safety.
5. Point formulas at cell ranges rather than numbers you typed by hand, so correcting one measurement updates every result that depends on it.
6. Round for display only and keep full precision in the cell, because rounding early is how totals stop adding up.

Choosing a Chart the Data Can Actually Support

1. Pick the chart from the question you are answering, not from the row of icons across the toolbar.
2. Categories with no natural order, such as colour choices, belong in a bar chart, because a line between them claims the gap means something.
3. Pie charts fall apart above about five slices, and they fail completely when the values are not parts of one whole.
4. Label both axes with the quantity and its unit, since "mm" and "score out of 5" are obvious to you and to nobody else.
5. One chart should make one point, and that point belongs under it as a full sentence.

Tables Belong in a Design Folder Where They Beat Prose

1. Reach for a table when a reader needs to compare the same properties across options, such as four materials rated on cost, mass, finish and workshop availability.
2. A specification test table with columns for specification point, test method, result and pass or fail is the fastest evidence you can give in Evaluating.
3. Keep tables short enough to read on one screen, because a 40 row table in a folder gets skipped and teaches nobody anything.
4. Put raw data in an appendix and the summary table in the body, so the argument stays readable while the evidence stays available.
5. Number and title every table and chart, then refer to it by that number when you discuss it.

Misleading Graphs: Small Choices That Change the Story

1. Starting a bar chart's vertical axis at 90 instead of 0 turns a 2 per cent gap into a bar that looks three times taller.
2. A cut axis is reasonable on a line graph tracking small changes over time, and dishonest on a bar chart where the reader compares heights.
3. Making one bar wider or adding 3D depth changes its apparent area, so it reads as larger than the number behind it.
4. Showing only the two test days that suit your argument and dropping the rest is cherry picking, and it shows up the moment somebody asks for the raw data.
5. A chart with no axis labels, no units and no sample size cannot be checked, so it should not convince you either.
6. Test your own charts by asking whether a reader who disagrees with your conclusion would still accept the picture.