Temperature compensation

All resistors change value somewhat when the temperature changes dramatically. And because resistors dissipate power, they can get hot just because of the current they carry. Often, this current is so tiny that it doesn’t appreciably heat the resistor. But in some cases it does, and the resistance might change. Then the circuit will behave differently than it did when the resistor was still cool.
There are various ways to approach problems of resistors changing value when they get hot.
One method is to use specially manufactured resistors that do not appreciably change value when they get hot. Such units are called temperature-compensated. But one of these can cost several times as much as an ordinary resistor.
Another approach is to use a power rating that is much higher than the actual dissipated power in the resistor. This will keep the resistor from getting very hot. Usually, it’s a needless expense to do this, but if the small change in value cannot be tolerated, it’s sometimes the most cost effective.
Still another scheme is to use a series-parallel network of resistors that are all identical, in the manner you already know about, to increase the power dissipation rating. Alternatively, you can take several resistors, say three of them, each with about three times the intended resistance, and connect them all in parallel. Or you can take several resistors, say four of them, each with about 1/4 the intended resistance, and connect them in series. It is unwise to combine several resistors with greatly different values. This can result in one of them taking most of the load while the others loaf, and the combination will be no better than the single hot resistor you started with.
You might get the idea of using two resistors with half (or twice) the value you need, but with opposite resistance-versus-temperature characteristics, and connecting them in series (or in parallel). Then the one whose resistance decreases with heat (negative temperature coefficient) will have a canceling-out effect on the one whose resistance goes up (positive temperature coefficient). This is an elegant theory, but in practice you probably won’t be able to find two such resistors without spending at least as much money as you would need to make a 3 ×3 series-parallel network. And you can’t be sure that the opposing effects will exactly balance. It would be better, in such a case, to make a 2 ×2 series-parallel array of ordinary resistors.