PlanMaker RSQ: Squared Correlation and Fitted-Line Variation

• Introduced: Documented by the PlanMaker 2010 manual; the first introduction release is not established.
• Availability: Documented in PlanMaker 2026; this is not a final working-version endpoint.
• Scope: The reference belongs to the 2026/NX manual family. Individual FreeOffice and mobile-build availability remains unverified.
• Checked: October 14, 2026, against the official function reference. Examples use vendor results and independent calculations; no native PlanMaker test was performed.

=RSQ(Range1, Range2)

Both arguments are required cell references. The PlanMaker reference says that the ranges should contain numbers; empty cells, text and logical values are ignored. Different numbers of values produce #N/A. It explicitly identifies the result as the square of PEARSON.

Keep the two measurements for each observation together. The short entry does not fully define alignment for differently positioned blanks or text, so the examples use complete numeric pairs.

The mathematical result lies from 0 to 1 for valid nonconstant data. A value of 0.6 can be displayed as 60%, but that is not a prediction accuracy or the percentage of individual responses the model gets right.

Create FitQuality. Enter this table in A1:B6. Leave C:E for the fit and I for summary formulas.

Enter the following summaries:

I4 and I5 describe the least-squares line y = 2.2 + 0.6*x. The equality of the slope and RSQ here is a coincidence: the slope has response-per-input units, while RSQ is unitless.

Put Fitted y, Squared residual, Squared deviation from mean in C1:E1. Enter =$I$5+$I$4*A2 in C2, =(B2-C2)^2 in D2 and =(B2-$I$6)^2 in E2. Fill C2:E2 through row 6.

In I8 enter =SUM(D2:D6): the residual sum of squares is 2.4. In I9 enter =SUM(E2:E6): the total squared variation about the mean is 6. Then I10, =1-I8/I9, gives 0.6.

For this ordinary least-squares line with a fitted intercept, 60% of the observed squared variation about the response mean is accounted for by the line. This interpretation is tied to that model and these observations. It does not give a confidence interval or certify future forecasts.

On FitQuality, put Negative y in K1, enter =-B2 in K2 and fill through K6.

Use PEARSON when the direction matters. Reversing the sign of the entire response column does not change its squared correlation.

In FitQuality, use a separate area O:Q to examine the line y = 1.2*x, constrained to pass through zero. Put Origin fit in O1 and Squared residual in P1. Enter =1.2*A2 in O2 and =(B2-O2)^2 in P2, then fill to row 6.

The fitted values are 1.2, 2.4, 3.6, 4.8 and 6. Put =SUM(P2:P6) in Q2; it gives 6.8. In Q3, =1-Q2/$I$9 gives approximately −0.133333333, although I2 still returns 0.6.

RSQ has not changed its definition. Squared correlation is not automatically the same goodness-of-fit calculation for a constrained line, arbitrary predictions or a different model.

Create VendorRSQ with 2, 5, 3 in A1:A3 and 2, 7, 4 in B1:B3. In D1 enter =RSQ(A1:A3,B1:B3). The independent result is 529/532, approximately 0.994360902, which rounds to the documented 0.99436. In D2, =PEARSON(A1:A3,B1:B3)^2 checks the same identity.

The documented mismatch error is #N/A. Constant inputs make the correlation denominator zero, but this page does not specify the exact native error token. Check the paired numeric records before interpreting an error or a surprisingly high result. Very small samples, unusual points and curved patterns can all make a single fit statistic misleading.

• PEARSON to retain the correlation’s sign.
• SLOPE to calculate the line’s rate of change.
• SUM to reconstruct squared-error totals.
• AVERAGE to calculate the response mean.