Charting intervals: what daily and weekly views each hide

Short answer#

A daily chart shows every entry, including the noise. A weekly chart shows direction, but it hides single spikes and any pattern that repeats inside the week. Neither one is the truth. Both are views of the same underlying entries, and the tracking interval you pick decides which story the chart tells.

Read your own log months later and that choice matters more than it seems. Below: what each interval shows, what it hides, and how to store your data so you can switch views without losing anything.

The interval is a choice#

Turning daily entries into weekly points is called temporal aggregation: converting observations from a fine interval into a coarse one. A 2014 paper on the modifiable temporal unit problem describes three separate effects of that choice: aggregation, segmentation and boundary.

• Aggregation is the size of the bucket: day, week or month.
• Segmentation is where the buckets start. The same series can be cut into different weekly segments depending on whether a week begins on Monday, Tuesday or Wednesday.
• Boundary is where the series starts and ends.

That paper studied crime events, not personal logs, so take it as a statement of the general concept only. In its data, daily, weekly and monthly views of the same events surfaced different clusters. The point carries over: the same entries can look different depending on how you slice them.

What a daily chart shows and hides#

A daily chart shows every entry you made, when you made it, and how far each sits from its neighbors. It is the right view for checking whether a value was logged correctly, whether an entry landed on the right day, and whether a single point looks like a typo.

What it hides is the trend. Day-to-day values move around for ordinary reasons, and a line that jumps up and down is hard to read as anything but jumps. The risk is reading one unusual day as a change. A single point is an observation, not a pattern, and a log is not a place to interpret it for yourself.

What a weekly chart shows and hides#

A weekly chart averages or sums the days inside each week. That makes direction easier to see, and it makes the chart calmer. It is a good view for asking whether the overall level has moved over several weeks.

It hides two things. First, short events: a one-day spike barely moves a seven-day average. Second, anything that repeats inside the week, which the next section covers.

| Interval | Shows | Hides | Good for | |---|---|---|---| | Daily | Every entry, outliers, entry timing | The underlying direction | Checking data quality | | Weekly | Direction over weeks | Single spikes, patterns inside the week | Looking at the longer trend | | Monthly | Slow, long-run change | Almost everything shorter | Seasonal or long-range review |

A weekly view can erase a weekly pattern#

Some data has a rhythm built into the week. In the NoHoW trial, 1,421 participants weighed themselves on connected smart scales. The authors found within-week fluctuations of about 0.35% (around 0.3 kg in that group), with weights highest around the weekend and lowest on Friday.

Two cautions. Everyone in that trial was in a weight-loss maintenance program, so the authors note the pattern may not represent the general population. And the point here is not about weight. It is that a repeating weekly pattern is visible on a daily chart and invisible on a weekly average, because the average folds the whole week into one number.

The opposite also holds. A simple moving average with a window equal to the period of a repeating fluctuation removes that fluctuation from the smoothed line. A seven-day window will flatten a seven-day rhythm. That is useful if you want the trend and misleading if the rhythm is the thing you are trying to see.

If you are tempted to line up a weekly pattern with something else in your log, read correlation in a personal log, and its limits first. Two things moving together on a smoothed chart is weak evidence.

Moving averages, smoothing and lag#

A moving average sits between a daily and a weekly view. It keeps one point per day but smooths each one using the days around it, which highlights longer-term movement and mutes short-term swings.

Two properties are worth knowing:

• Lag. A trailing simple moving average lags the latest entry by half the window width. A seven-day average reflects roughly where things stood three or four days ago.
• Window length. A longer window gives a smoother line and a larger delay. The right window depends on the kind of movement you want to see.

The same source notes that a simple moving average can let through some signal shorter than its window, which can put peaks in the smoothed line where the raw data had troughs. Keep the raw points visible behind the smoothed line so you can compare the two. The same idea is covered for body measurements in weight trends versus weight readings.

Gaps and missing days#

Each chart interval handles missing days differently. A daily chart shows a hole. A weekly average quietly computes from whatever days exist, so a week with two entries looks just as solid as a week with seven.

Two rules help:

1. Do not fill gaps silently. The NoHoW authors chose not to impute missing days because imputation reduces the differences between sequential days, which removes some real variability. If you smooth over a gap, mark it.
2. Count the gap itself as data. In a Finnish dataset of 40 adults and 2,838 weight entries, the change between two entries was associated with the time between them, and the authors flagged breaks longer than a month. It was a small, homogeneous sample, so read it as an illustration. A log that records how long it went unchecked is more honest than one that hides it.

When you miss an entry, write that down instead of leaving it blank. Missed dose log: what to record when you miss an entry covers how.

What to record#

A few habits keep your options open, whatever interval you chart at:

• Store the raw, timestamped entries. Daily, weekly and monthly views can all be rebuilt from them. A weekly average cannot be turned back into daily values.
• Label the interval and the week start. A chart that says "weekly" without a start day cannot be compared with another one.
• Mark gaps with a visible note rather than a smoothed line.
• Check a second interval before you decide something has changed. If a shift shows up on the daily and the weekly view, it is more likely real than one that shows up on only one.
• Keep a copy outside the app. If you ever need to rebuild a chart, you want the raw entries to be exportable. See backing up and exporting your log.

OptimizedTracker stores each entry with its own timestamp, so the daily and weekly views always come from the same raw record.

Conclusion#

Choose the interval for the question you are asking. Use daily to check data quality and to see rhythm inside the week. Use weekly or a moving average to see direction. Before you decide anything has shifted, look at it at a second interval, and keep the raw entries so you always can.